Energy efficiency optimization method for box-type substation

Abstract

The present application relates to the technical field of electric energy distribution and power supply control, and particularly relates to an energy efficiency optimization method of a box-type substation. The method acquires operation data of a transformer, an incoming-outgoing line loop, a reactive power compensation device, a station power load and a temperature control device, and constructs an electrical-thermal coupling state diagram representing the relationship of electric energy transmission, reactive power flow, heat transfer and auxiliary power consumption; based on the state diagram, the loss contribution of each device and the control action propagation are calculated, a candidate coordinated control strategy including reactive power compensation switching, temperature control starting and stopping and loop operation constraint adjustment is generated, and a target coordinated control strategy with smaller comprehensive loss is determined under the constraints of voltage, current-carrying capacity, temperature and action interval. The method makes the energy consumption of multiple devices in the station associated in a unified model to participate in control decision, and reduces the problems of high comprehensive loss and repeated action caused by local control.

Description

Technical Field

[0001] This invention relates to the field of power distribution and power supply control technology, specifically to a method for optimizing the energy efficiency of a prefabricated substation. Background Technology

[0002] A prefabricated substation typically consists of a transformer, incoming circuits, outgoing circuits, reactive power compensation equipment, station auxiliary loads, and temperature control equipment. It is used to complete the access, transformation, distribution, and operational protection of electrical energy on the distribution side. Current energy efficiency control largely relies on in-station monitoring devices to collect data such as voltage, current, active power, reactive power, power factor, transformer load rate, and temperature. These data are then used by monitoring terminals or distribution automation platforms to make judgments based on preset thresholds. In conventional schemes, reactive power compensation equipment is switched on and off based on changes in power factor or reactive power; temperature control equipment is started and stopped based on the temperature inside the substation or the transformer temperature; circuit operating status is monitored based on current, voltage, and protection settings; and station auxiliary loads are statistically analyzed separately as auxiliary power consumption. The above control methods treat single types of equipment or single indicators as processing objects, with clear control logic boundaries and simple engineering implementation. They are suitable for ensuring basic operational requirements such as qualified voltage, no equipment overload, and no temperature exceeding limits.

[0003] In the closest existing technologies, substation energy efficiency optimization is typically achieved through statistical analysis of individual losses and local control corrections. Transformer losses are estimated based on load rate or historical efficiency curves, line loop losses are calculated based on current and loop impedance, reactive power compensation control determines capacitor bank switching based on power factor deviation, and temperature control equipment starts and stops according to temperature upper and lower limits or hysteresis ranges. Some solutions aggregate the above data into an energy efficiency analysis platform, generating energy consumption reports, loss rankings, or alarm information, and then issue equipment control commands according to preset control priorities. While this type of solution can track the operating status of individual devices and trigger corresponding actions when local indicators are abnormal, the relationships between power transmission, reactive power flow, heat transfer, and auxiliary power consumption between devices are usually not uniformly expressed, and the mutual influence between equipment control actions lacks quantitative processing under a common calculation framework.

[0004] The main technical problem with existing technologies is that the energy consumption states of multiple devices in a prefabricated substation have electrical and thermal coupling relationships. Conventional control methods still rely mainly on independent thresholds, fixed priorities, or individual loss statistics, making it difficult to determine the propagation impact of a single control action on the overall station loss within the same control cycle. Reactive power compensation switching may change the reactive power flow and voltage level of the circuit; temperature control start-up and shutdown will change the station's power consumption and affect thermal state changes; and circuit operation constraint adjustments will change the load distribution and transformer load losses. These effects are handled in a decentralized manner in local control strategies, resulting in a lack of a unified basis for comparison among candidate control strategies. When individual indicators are within limits but the overall loss is high, existing solutions struggle to determine the equipment and action combinations that should be prioritized for adjustment, easily leading to control results where local energy efficiency improvements coexist with overall station energy efficiency deterioration. Summary of the Invention

[0005] The purpose of this invention is to provide an energy efficiency optimization method for prefabricated substations, which can effectively solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An energy efficiency optimization method for a prefabricated substation includes: acquiring operating data of transformers, incoming and outgoing circuits, reactive power compensation equipment, station power loads and temperature control equipment within the prefabricated substation; and constructing an electrical-thermal coupling state diagram based on the operating data to characterize the relationships of power transmission, reactive power flow, heat transfer and auxiliary power consumption. Calculate the loss contribution and control action propagation of each device based on the electrical-thermal coupling state diagram. Generate candidate coordinated control strategies that include reactive power compensation switching, temperature control start-stop, and loop operation constraint adjustment; A method for determining and executing a target collaborative control strategy with the minimum overall loss from the candidate collaborative control strategies, under the conditions of satisfying voltage, current carrying capacity, temperature, and action interval constraints.

[0007] Preferably, the construction of the electrical-thermal coupling state diagram includes: mapping the transformer, the incoming and outgoing line circuits, the reactive power compensation equipment, the station power load, and the temperature control equipment as state nodes respectively; The active power transmission relationship between the transformer and the incoming and outgoing line circuits, the reactive power adjustment relationship between the reactive power compensation device and the incoming and outgoing line circuits, the heat load transfer relationship between the transformer and the temperature control device, and the auxiliary power consumption relationship between the temperature control device and the station power load are respectively mapped as associated edges; The weights of the associated edges are updated based on the voltage, current, active power, reactive power, load rate, and temperature data within the same sampling period.

[0008] Preferably, the calculation of the loss contribution includes: according to the state node type in the electrical-thermal coupling state diagram, the operating data is divided into transformer load loss data, circuit transmission loss data, reactive power flow loss data, station power consumption data, and temperature control power consumption data; Normalize and map the transformer load loss data, the circuit transmission loss data, the reactive power flow loss data, the station power consumption data, and the temperature control power consumption data within the same control cycle; Based on the weights of the associated edges, the normalized data is coupled and superimposed to obtain the loss contribution of each state node.

[0009] Preferably, the calculation of the control action propagation includes: establishing action vectors for the reactive power compensation switching, the temperature control start / stop, and the loop operation constraint adjustment, respectively; The motion vector is input into the electrical-thermal coupling state diagram, and the calculation is performed layer by layer along the reactive power flow relationship, the heat transfer relationship, and the auxiliary power consumption relationship. The voltage change, reactive power change, temperature change, and station power consumption change after propagation are jointly encoded to obtain the control action propagation quantity of the corresponding candidate collaborative control strategy; In this context, action vectors with conflict propagation directions within the same candidate collaborative control strategy are merged into a single action combination.

[0010] Preferably, the generation of the candidate collaborative control strategy includes: selecting state nodes whose loss contribution meets the preset contribution conditions as nodes to be adjusted, based on the loss contribution amount in the current control cycle as the sorting basis; The set of cooperative devices is determined based on the adjacent associated edges of the node to be adjusted in the electrical-thermal coupling state diagram; Within the set of collaborative devices, generate reactive power compensation switching sequences, temperature control start / stop sequences, and loop operation constraint adjustment sequences; The reactive power compensation switching sequence, the temperature control start-stop sequence, and the loop operation constraint adjustment sequence are combined to form the candidate collaborative control strategy, and combinations that do not meet the equipment action interval constraints are eliminated.

[0011] Preferably, the weight update of the associated edge includes: writing the changes in active power, reactive power and temperature in adjacent sampling periods into the dynamic weight items of the corresponding associated edge; Write the transformer load rate, the switching status of the reactive power compensation equipment, and the start / stop status of the temperature control equipment into the state weight item of the corresponding associated edge; A weight correction factor is generated according to the deviation direction between the dynamic weight term and the state weight term of the same associated edge; The weights of the electrical energy transmission relationship, the reactive power flow relationship, the heat transfer relationship, and the auxiliary power consumption relationship are updated synchronously using the weight correction factor.

[0012] Preferably, the coupling superposition includes: establishing a nonlinear loss mapping term corresponding to the load rate for the transformer load loss data; A reactive power deviation mapping term corresponding to the power factor offset direction is established for the aforementioned reactive power flow loss data; A thermal inertia mapping term corresponding to the temperature change hysteresis is established for the aforementioned temperature control power consumption data; The nonlinear loss mapping term, the reactive power deviation mapping term, and the thermal inertia mapping term are multiplied by the weights of their respective associated edges and then input into the same loss contribution matrix. The loss contribution of each state node is then obtained based on the loss contribution matrix.

[0013] Preferably, the merging of action vectors with conflict propagation direction includes: identifying a first action vector in the same candidate cooperative control strategy that reduces reactive power change and increases station power consumption change, and a second action vector that reduces temperature change and increases station power consumption change. Generate an action priority order based on the loss contribution, remaining action interval time, and propagation level corresponding to the first action vector and the second action vector; Based on the action priority, the dominant action vector is retained, and the non-dominant action vectors are converted into delayed action vectors, forming a single action combination with a time sequence.

[0014] Preferably, after the weights of the associated edges are updated synchronously, the method further includes: obtaining the target collaborative control strategy of the previous control cycle and the corresponding execution data, and writing the difference in loss contribution before and after execution, the difference in control action propagation, and the difference in associated edge weights into the action memory table. When generating candidate collaborative control strategies for the current control cycle, the action cost is increased based on the action memory table for action combinations that are repeated and whose difference in loss contribution does not meet the convergence condition. The action cost is incorporated into the overall loss calculation process to obtain an evaluation value used to screen target collaborative control strategies.

[0015] Preferably, after forming the single action combination with time sequence, the method further includes: inputting the single action combination into a constraint verification sequence consisting of voltage constraint, current constraint, temperature constraint, reactive power compensation action interval constraint and temperature control start-stop hold constraint; When the single action combination fails the constraint verification sequence, the delayed action vector is replaced or the non-dominant action vector is deleted in sequence according to the action priority, and a new candidate cooperative control strategy is formed. When there are multiple candidate cooperative control strategies that pass the constraint verification sequence, the candidate cooperative control strategy with the smallest evaluation value is selected as the target cooperative control strategy.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By mapping transformers, incoming and outgoing circuits, reactive power compensation equipment, station power loads, and temperature control equipment as state nodes, and mapping the relationships between power transmission, reactive power flow, heat transfer, and auxiliary power consumption as associated edges, the energy consumption relationships among multiple devices within a prefabricated substation can be expressed in the same electrical-thermal coupled state diagram. After calculating the loss contribution and control action propagation based on operational data, reactive power compensation switching, temperature control start / stop, and circuit operation constraint adjustments are no longer judged as isolated actions, but are combined, verified, and screened within the same candidate collaborative control strategy. Therefore, under constraints of voltage, current carrying capacity, temperature, and action interval, the comprehensive loss can be directly used as the comparison object for control strategies, avoiding local control deviations caused by triggering only a single threshold, and ensuring that the target collaborative control strategy remains consistent with the overall station energy efficiency status.

[0017] 2. By dynamically updating the weights of associated edges and incorporating changes in active power, reactive power, temperature, transformer load rate, reactive power compensation switching status, and temperature control start / stop status into the weight correction process, the electrical-thermal coupling state diagram can be adjusted to adapt its calculation basis according to changes in operating conditions. By processing different types of loss data through nonlinear loss mapping, reactive power deviation mapping, and thermal inertia mapping, the bias of single linear loss estimation for complex operating conditions can be reduced. By identifying the action vectors in the conflict propagation direction and forming a combination of single actions with a temporal sequence, the cancellation of control actions within the same cycle can be reduced. By recording the difference in loss contribution, control action propagation amount, and associated edge weight before and after execution using an action memory table, and incorporating the action cost into the evaluation process, the probability of repetitive, unprofitable actions entering the candidate strategy can be reduced, making the continuous control process more stable. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the overall method of the present invention; Figure 2 This is a flowchart illustrating the construction of the electrical-thermal coupling state diagram of the present invention; Figure 3 Flowchart for calculating the loss contribution and control action propagation amount of the present invention; Figure 4 This is a flowchart of the candidate collaborative control strategy generation and constraint screening process of the present invention; Figure 5 This is a flowchart of the motion memory table and rolling closed-loop control of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please refer to Figure 1 This embodiment provides an energy efficiency optimization method for prefabricated substations, applied to an existing monitoring and control platform within the substation. The processing objects include transformers, incoming circuits, outgoing circuits, reactive power compensation equipment, station power loads, and temperature control equipment. The method does not change the hardware structure of the prefabricated substation and does not add any data acquisition objects unrelated to energy efficiency optimization. The operational data used comes from the substation's electrical quantity acquisition, equipment status acquisition, and temperature acquisition results. The operational data includes voltage, current, active power, reactive power, power factor, transformer load rate, internal temperature, transformer temperature, reactive power compensation equipment switching status, temperature control equipment start / stop status, and station power load operating status. The control platform performs time alignment on the above operational data according to a unified sampling period, removing missing timestamps, duplicate timestamps, and data records that cross physical value boundaries. Data belonging to the same equipment or the same circuit within the same sampling period is encapsulated into a state vector. For incoming and outgoing circuits, the state vector includes at least voltage, current, active power, and reactive power; for transformers, the state vector includes at least load rate, input power, output power, and temperature; for reactive power compensation equipment, the state vector includes at least switching status, reactive power changes before and after switching, and corresponding circuit voltage changes; for temperature control equipment, the state vector includes at least start / stop status, internal temperature changes, and station power consumption changes. This processing method allows electrical and thermal states within the same control cycle to be included in the same calculation process, avoiding subsequent control strategies based on data comparisons across different time windows. The advantage of this embodiment is that the energy efficiency calculation basis remains consistent with the real-time operating status of multiple devices within the station, providing a unified data entry point for subsequent collaborative control.

[0021] Specifically, refer to Figure 2In one embodiment, the control platform constructs an electrical-thermal coupling state diagram based on preprocessed operational data. This diagram consists of state nodes and associated edges. State nodes represent the operating states of transformers, incoming and outgoing circuits, reactive power compensation equipment, station power loads, and temperature control equipment. Associated edges represent relationships such as power transmission, reactive power flow, heat transfer, and auxiliary power consumption. An input power transmission edge is established between the transformer node and the incoming circuit node; an output power transmission edge is established between the transformer node and the outgoing circuit node; a reactive power flow edge is established between the reactive power compensation equipment node and its corresponding incoming and outgoing circuit node; a heat transfer edge is established between the transformer node and the temperature control equipment node; and an auxiliary power consumption edge is established between the temperature control equipment node and the station power load node. Each associated edge is configured with edge type, edge direction, real-time weight, and historical weight. The edge direction does not represent a unidirectional constraint of the physical connection structure but is used to determine the transmission order of state changes when calculating the propagation amount of control actions. Real-time weights are determined by the electrical quantities, heat, and equipment status of the corresponding device or circuit within the current sampling period, while historical weights are formed by smoothing a weight sequence from multiple consecutive sampling periods. Table 1 shows the correspondence between state nodes, data sources, and state meanings, illustrating the data organization method of the electrical-thermal coupling state diagram.

[0022] Table 1. Correspondence between status nodes and running data

[0023] In this embodiment, the real-time weight of each associated edge is calculated using standardized state differences. For power transmission edges, the weight is related to changes in active power and current; for reactive power flow edges, the weight is related to changes in reactive power and power factor shift; for heat transfer edges, the weight is related to temperature changes and transformer load rate; and for auxiliary power consumption edges, the weight is related to the start / stop status of temperature control equipment and changes in station power consumption. To ensure that data of different dimensions enter the same graph model, the control platform normalizes each piece of operational data using historical operational intervals, and the normalized values ​​are used in the associated edge weight calculation. The real-time weight of the associated edge between the i-th state node and the j-th state node in the t-th sampling period is calculated as follows: ; in, Indicates the first The state node to the _th Real-time weights of the edges associated with each state node; This represents the normalized change in active power. This represents the normalized change in reactive power. This represents the normalized change in temperature. This represents the normalized change in equipment status; , , , These represent the weight coefficients corresponding to power transmission, reactive power flow, heat transfer, and auxiliary power consumption, respectively, with the sum of all weight coefficients for the same associated edge being 1. For example, if a heat transfer edge has a normalized change in active power of 0.3, a normalized change in reactive power of 0.1, a normalized change in temperature of 0.6, and a normalized change in equipment status of 0.2 in a certain sampling period, with corresponding coefficients of 0.2, 0.1, 0.6, and 0.1 respectively, then its real-time weight is 0.42. This result indicates that within the current sampling period, this associated edge is mainly driven by temperature changes and transformer load changes. The advantage of this embodiment is that the power, reactive, heat, and station power consumption relationships are compressed into the same graph weight expression, facilitating subsequent unified calculation of loss contributions and control action propagation.

[0024] In one embodiment, reference Figure 3 The control platform calculates the loss contribution of each device based on the electrical-thermal coupling state diagram. This loss contribution is not a static loss estimate of a single device, but rather a calculation result formed under the combined effect of the current state node's own loss data and the coupling influence of adjacent associated edges. For transformer nodes, the control platform establishes transformer load loss data based on load rate, input-side active power, output-side active power, and temperature; for incoming and outgoing line loop nodes, it establishes loop transmission loss data based on loop current, active power, and reactive power; for reactive power compensation equipment nodes, it establishes reactive power flow loss data based on reactive power changes and voltage changes before and after switching; for station power load nodes, it establishes station power consumption data based on auxiliary equipment power consumption; and for temperature control equipment nodes, it establishes temperature control power consumption data based on temperature control start / stop status, temperature changes, and station power consumption changes. All of the above loss data are entered into the same loss contribution matrix, where rows represent state nodes and columns represent loss types. Each matrix element is normalized and coupled and superimposed according to the weights of the corresponding associated edges. This is used to calculate the loss contribution of the i-th state node in the t-th sampling period. ; in, Indicates the first The loss contribution of each state node; Indicates the first The local loss normalization value of each state node; Indicates the relationship with the first A set of adjacent state nodes for each state node that has an associated edge; Indicates the first The state node to the _th Real-time weights of the edges associated with each state node; Indicates the first The local loss normalization value of each state node. For example, the local loss normalization value of a temperature control device node is 0.25, the local loss normalization value of the adjacent transformer node is 0.70, the weight of the heat transfer edge between them is 0.40, the local loss normalization value of the adjacent station power load node is 0.30, and the auxiliary power consumption edge weight is 0.20. Then, the loss contribution of the temperature control device node is 0.59. This calculation method can simultaneously include the power consumption of the temperature control device itself, the heat load transfer of the transformer, and the station power consumption association into a single node contribution value. The advantage of this embodiment is that the control platform can identify nodes with a high contribution to the overall station loss, rather than judging solely based on the power consumption of a single device or the exceeding of a single limit.

[0025] In this embodiment, the control platform further calculates the propagation amount of control actions. This propagation amount describes the impact of reactive power compensation switching, temperature control start-up / stop, and loop operation constraint adjustment on other nodes in the electrical-thermal coupling state diagram. The control platform establishes an action vector for each type of control action. The action vector includes at least the action object, action direction, action occurrence period, action duration period, and action state change. The state change of the reactive power compensation switching action vector consists of reactive power change and voltage change; the state change of the temperature control start-up / stop action vector consists of station power consumption change and temperature change; and the state change of the loop operation constraint adjustment action vector consists of loop power change, loop current change, and transformer load rate change. After inputting the action vector into the electrical-thermal coupling state diagram, the control platform propagates it layer by layer along the reactive power flow edge, heat transfer edge, and auxiliary power consumption edge, and attenuates the state change at each propagation level. The propagation amount of action vector r at the k-th propagation level is used to calculate: ; in, Represents action vector In the The amount of transmission at each level; Indicates the first The set of associated edges contained in each propagation level; Represents action vector Acting on the The state change amount when there are 1 state node; This represents the propagation attenuation coefficient, with a value greater than 0 and less than 1. Indicates the first The state node to the _th The real-time weights of the associated edges of each state node. For example, if a reactive power compensation switching action generates a reactive power change of 0.50 at the starting node, and the weight of the first-layer reactive power flow edge is 0.60 with a propagation attenuation coefficient of 0.80, then the first-layer propagation amount is 0.24; if the weight of the second-layer associated edge is 0.40, then the second-layer propagation amount is 0.064. By calculating the propagation layer by layer, the comprehensive impact of this action on voltage changes, reactive power changes, station power consumption changes, and temperature changes can be obtained. The advantage of this embodiment is that it can quantify the transmission results of actions between equipment within the station before the control strategy is generated, reducing the situation where only the state of the action object itself is considered while ignoring the state of associated equipment.

[0026] In one embodiment, reference Figure 4 The control platform generates candidate coordinated control strategies based on loss contribution and control action propagation. The generation process uses the loss contribution of the current control cycle as a basis, selecting state nodes whose loss contribution meets preset contribution conditions as nodes to be adjusted. These preset contribution conditions can be ranking conditions or relative contribution conditions, such as selecting nodes with high loss contributions or nodes whose loss contribution is higher than the average contribution of all nodes in the station. The control platform determines the set of cooperative devices based on the adjacent edges of the node to be adjusted in the electrical-thermal coupling state diagram. If the node to be adjusted is a reactive power compensation device node, the set of cooperative devices includes incoming and outgoing line loop nodes with reactive power flow edges and transformer nodes associated with those loops; if the node to be adjusted is a temperature control device node, the set of cooperative devices includes transformer nodes with heat transfer edges and station power load nodes with auxiliary power consumption edges; if the node to be adjusted is an outgoing line loop node, the set of cooperative devices includes related incoming line loop nodes, transformer nodes, and reactive power compensation device nodes. Subsequently, the control platform generates reactive power compensation switching sequences, temperature control start-stop sequences, and loop operation constraint adjustment sequences within the set of cooperative devices, and combines these sequences to form candidate cooperative control strategies. Combinations that do not meet the device action interval constraints are eliminated by the control platform before entering the evaluation process. The advantage of this embodiment is that the candidate strategies originate from high-loss contributing nodes and their coupled devices, reducing the entry of irrelevant actions into the strategy space.

[0027] Preferably, when evaluating candidate collaborative control strategies, the control platform not only calculates the comprehensive loss after the action is completed, but also performs unified verification of the constraints. The constraints include voltage constraints, current-carrying constraints, temperature constraints, and action interval constraints. Voltage constraints are used to limit the voltage of each incoming and outgoing line circuit to within the allowable operating range; current-carrying constraints are used to limit the current of each circuit to not exceed the allowable current-carrying boundary; temperature constraints are used to limit the temperature inside the enclosure and the transformer temperature to within the allowable operating range; and action interval constraints are used to limit the reactive power compensation equipment and temperature control equipment from repeating actions within a short period. The evaluation value of a candidate strategy is composed of the predicted comprehensive loss value and the action cost. The predicted comprehensive loss value is obtained by adding the predicted loss contribution values ​​of each node after the candidate strategy is executed. The action cost is determined by the action type, the remaining time of the action interval, whether the action is repeated, and the degree of action propagation conflict. The evaluation value of candidate collaborative control strategy u in the t-th control cycle is calculated as follows: ; in, Indicates candidate collaborative control strategies Evaluation value; Indicates the number of state nodes; Indicates candidate collaborative control strategies After execution Predicted loss contribution of each state node; Indicates candidate collaborative control strategies The cost of the action; Indicates candidate collaborative control strategies The constraint deviation penalty value; Indicates the action cost coefficient; This represents the constraint deviation penalty coefficient. For example, if the sum of the predicted loss contributions of all nodes after the execution of a candidate strategy is 2.80, the action cost is 0.30, the constraint deviation penalty value is 0, the action cost coefficient is 0.50, and the constraint deviation penalty coefficient is 1, then the evaluation value is 2.95. If the sum of the predicted loss contributions of another candidate strategy is 2.70, but the constraint deviation penalty value is 0.40, then the evaluation value is 3.10 under the same coefficient, and the former candidate strategy is selected first. The advantage of this embodiment is that the determination of the target collaborative control strategy simultaneously considers loss, action continuity, and operational constraints, and the control result is consistent with the overall station operation boundary.

[0028] In a preferred embodiment, the weight update of the associated edge is jointly corrected by dynamic weight terms and state weight terms. The dynamic weight terms are formed by the changes in active power, reactive power, and temperature within adjacent sampling periods, while the state weight terms are formed by the transformer load rate, the switching status of reactive power compensation equipment, and the start / stop status of temperature control equipment. For the power transmission edge, the dynamic weight terms mainly come from the changes in active power and current, while the state weight terms mainly come from the transformer load rate. For the reactive power flow edge, the dynamic weight terms mainly come from the changes in reactive power and power factor, while the state weight terms mainly come from the switching status of reactive power compensation equipment. For the heat transfer edge, the dynamic weight terms mainly come from the temperature changes, while the state weight terms mainly come from the transformer load rate and the start / stop status of temperature control equipment. For the auxiliary power consumption edge, the dynamic weight terms mainly come from the changes in station power consumption, while the state weight terms mainly come from the start / stop status of temperature control equipment. The control platform compares the deviation direction of the dynamic weight term and the state weight term of the same associated edge. If the two change in the same direction, the participation weight of the associated edge in the next control cycle is increased; if the two change in opposite directions, the participation weight of the associated edge in the next control cycle is decreased. Table 2 shows the basis for weight updates of different associated edges, which is used to explain the input data and correction objects of various types of edges.

[0029] Table 2. Basis for updating the weights of associated edges

[0030] Furthermore, in this embodiment, the weight correction adopts a smooth update method to avoid abnormal fluctuations in a single sampling period directly changing the subsequent control direction. This is used to calculate the associated edge between the i-th state node and the j-th state node in the... Update weights for each sampling period: ; in, This represents the updated weight of the associated edges; This represents the weight of the associated edges in the current sampling period; Indicates a dynamic weight term; Represents the state weight term; The smoothing coefficient represents a value greater than 0 and less than 1. For example, if the current weight of a reactive power flow edge is 0.50, the dynamic weight is 0.70, the state weight is 0.60, and the smoothing coefficient is 0.60, then the updated weight will be 0.56. This update retains a portion of the current weights while incorporating changes in the current operating state. The advantage of this embodiment is that the electrical-thermal coupling state diagram can adjust its calculation basis according to changes in the operating state, while avoiding frequent switching of candidate strategies caused by short-term noise.

[0031] In a preferred embodiment, the coupled superposition of loss contributions further introduces nonlinear loss mapping terms, reactive power deviation mapping terms, and thermal inertia mapping terms. The relationship between transformer load loss and load rate is not a simple linear one; when the load rate deviates from the economic operating range, the load loss contribution exhibits a nonlinear upward trend with load rate changes. Therefore, the control platform establishes a nonlinear loss mapping term corresponding to the load rate for the transformer load loss data. Reactive power flow loss is related to the power factor offset direction. When the switching of reactive power compensation equipment causes overcompensation or undercompensation, the reactive power deviation mapping term is assigned different calculation symbols according to the offset direction. Temperature control power consumption has a time lag with temperature changes. When the temperature control equipment has started but the temperature inside the tank has not yet decreased, the thermal inertia mapping term is calculated based on the difference between the current temperature change and the previous temperature change. The control platform multiplies the above three types of mapping terms by their corresponding associated edge weights and writes them into the loss contribution matrix, ensuring that data with different physical attributes retain differentiated calculation logic in the same matrix. Table 3 shows the composition of the loss mapping terms, illustrating the source of each mapping term in the loss contribution matrix.

[0032] Table 3. Composition of Loss Mapping Term

[0033] In this embodiment, the nonlinear loss mapping term, reactive power deviation mapping term, and thermal inertia mapping term do not replace the basic loss data, but rather serve as corrections to the basic loss data. If the load rate of a transformer node continuously deviates from the economic operating range, the nonlinear loss mapping term increases the weight of that node in the loss contribution matrix. If the power factor shift direction reverses after a reactive power compensation device is switched on, the reactive power deviation mapping term changes the contribution direction of the corresponding reactive power flow edge in the next control cycle. If the temperature drop lags after a temperature control device is started, the thermal inertia mapping term increases the proportion of the correlation calculation between the temperature control device node and the station power load node. The resulting loss contribution matrix can simultaneously express electrical losses, reactive power deviation, station power consumption, and thermal state lag. The advantage of this embodiment is that the loss contribution calculation can reflect the common nonlinear load, reactive power deviation, and thermal inertia phenomena in prefabricated substations, reducing the judgment bias caused by single linear loss estimation.

[0034] In one embodiment, the control platform identifies and merges action vectors with conflict propagation directions. A conflict propagation direction refers to a situation within the same candidate collaborative control strategy where an action reduces one type of loss contribution but simultaneously increases another through an associated edge. For example, reactive power compensation switching actions may reduce reactive power flow losses, but frequent switching increases the cost of the action and causes voltage fluctuations; temperature control start-stop actions may reduce temperature-related loss contributions but simultaneously increase station power consumption; loop operation constraint adjustments may reduce transmission losses in a certain loop but increase the load contribution of other loops or transformer nodes. After candidate strategies are formed, the control platform encodes the propagation result of each action vector in terms of direction. Direction encoding includes reactive power change direction, voltage change direction, temperature change direction, station power consumption change direction, and load change direction. If two action vectors exhibit opposite direction codes on the same associated edge or adjacent propagation levels, and the opposite directions involve major contributors in the comprehensive loss evaluation value, then the two action vectors are determined to have conflict propagation directions. For conflicting actions, the control platform does not directly delete them but generates an action priority order based on loss contribution, remaining action interval time, and propagation level. Actions with high loss contribution, actions whose action interval constraints are met, and actions with short propagation levels and clear impact paths are given higher priority; other actions are converted into delayed action vectors, which together with the dominant action vector form a single action combination with a time sequence. The advantage of this embodiment is that action combinations within the same control cycle can undergo conflict direction processing before entering constraint verification, reducing mutually canceling actions coexisting in the target cooperative control strategy.

[0035] Further, in a preferred embodiment, the control platform executes a constraint verification sequence for single action combinations with a time sequence. The constraint verification sequence consists of voltage constraints, current-carrying constraints, temperature constraints, reactive power compensation action interval constraints, and temperature control start-stop hold constraints. Voltage constraints determine whether the predicted voltage values ​​of each circuit are within the allowable operating range after reactive power compensation switching or circuit operation constraint adjustment; current-carrying constraints determine whether the predicted circuit current values ​​are within the allowable current-carrying range; temperature constraints determine whether the predicted temperatures inside the enclosure and the transformer are within the allowable operating range; reactive power compensation action interval constraints determine whether the time interval between adjacent switching of the same reactive power compensation device meets the device action records stored in the control platform; and temperature control start-stop hold constraints determine whether the start-stop state of the temperature control device meets the minimum hold time. If a single action combination fails the constraint verification sequence, the control platform processes delayed action vectors according to action priority, prioritizing the replacement of lower-priority delayed action vectors; if the constraint verification still fails after replacement, non-dominant action vectors are deleted and new candidate strategies are formed. If multiple candidate strategies pass the constraint verification sequence, a target collaborative control strategy is selected based on the evaluation value. The advantage of this embodiment is that the control strategy undergoes electrical safety, thermal safety, and operation continuity verification before being issued, which can prevent strategies with low overall losses but unmet operating boundaries from being executed.

[0036] In one embodiment, reference Figure 5 The control platform sets up an action memory table to record the target collaborative control strategy of the previous control cycle and its operational changes after execution. The action memory table includes action combination identifiers, pre-execution loss contribution, post-execution loss contribution, pre-execution control action propagation amount, post-execution control action propagation amount, pre-execution associated edge weights, post-execution associated edge weights, and action cost. The action combination identifier is generated based on the action object, action direction, and action occurrence cycle, and is used to determine whether the current candidate strategy is the same as or similar to historical action combinations. The difference in loss contribution before and after execution describes the change in loss contribution of each state node after the action is executed; the difference in control action propagation amount describes whether the action propagation path is consistent with the prediction; and the difference in associated edge weights describes whether the electrical-thermal coupling state diagram has undergone stable updates due to the action. If a certain action combination occurs repeatedly, and the difference in loss contribution before and after execution does not meet the convergence condition, the control platform adds an action cost to that action combination in the current cycle. The action cost is used to calculate the action cost of the candidate collaborative control strategy u. ; in, Indicates candidate collaborative control strategies In the The action value of each control cycle; This represents the action interval constraint; the closer the action interval is to the limit boundary, the larger the value. This indicates a historical recurrence; the closer the candidate strategy is to the non-converged action combination in the action memory table, the larger the value. This represents the action propagation conflict term; the more conflict propagation directions there are between action vectors, the larger the value. , , These represent the calculated coefficients for the three categories, and the sum of these coefficients is 1. For example, if a candidate strategy has an action interval constraint of 0.40, a historical repetition factor of 0.60, and an action propagation conflict factor of 0.20, with coefficients of 0.30, 0.50, and 0.20 respectively, then the action cost value is 0.46. This action cost value is subsequently incorporated into the candidate strategy evaluation process. The advantage of this embodiment is that the continuous control process can refer to historical execution results, reducing the probability of repeated, unprofitable actions being selected in subsequent cycles.

[0037] In this embodiment, the target collaborative control strategy is executed using a rolling control method. In each control cycle, the control platform only issues the reactive power compensation switching command, temperature control start / stop command, or loop operation constraint adjustment command that should be executed in the current cycle, while retaining the right to recalculate in the next cycle. After the target collaborative control strategy is issued, the control platform continues to collect operating data for the next sampling cycle, comparing the executed voltage, current, active power, reactive power, temperature, and equipment status with the predicted values ​​before execution. If the direction of data change after execution is consistent with the predicted direction of the control action propagation, the current associated edge weight update result is retained; if the direction of data change after execution is opposite to the predicted direction, the priority of the corresponding action vector in subsequent candidate strategies is reduced, and the dynamic weights of the relevant associated edges are adjusted. For strategy replacement records caused by voltage constraints, current-carrying constraints, or temperature constraints, the control platform writes the replacement reason into the action memory table, ensuring that the next control cycle avoids combinations of the same type that do not meet the constraints when generating candidate strategies. The execution records of the target collaborative control strategy and the action memory table together form closed-loop data, ensuring that each control action can become input for subsequent strategy selection. The advantage of this embodiment is that the control process can be corrected according to continuously sampled data, avoiding deviations in the operating state caused by using the same control combination for a long time.

[0038] In a preferred embodiment, the control platform sets consistency checks on the input operating data to improve the computational reliability of the electrical-thermal coupled state diagram. Consistency checks include time consistency, electrical consistency, and state consistency. Time consistency is used to determine whether various types of data are within the same sampling period. If the timestamps of the electrical data and state data of the same device are inconsistent, the control platform generates a state vector based on the most recent complete sampling period. Electrical consistency is used to determine whether the incoming active power, outgoing active power, station power consumption, and transformer losses are within an interpretable range. If the differences are too large, the relevant state nodes are marked as low-confidence nodes, and their participation weight in candidate strategy generation is reduced. State consistency is used to determine whether changes in device state match changes in corresponding electrical quantities. For example, after a change in the switching state of reactive power compensation equipment, the reactive power of the corresponding circuit should change in the same direction; after a change in the start / stop state of temperature control equipment, the station power consumption should change in the same direction. If the state change and the electrical quantity change do not match, the control platform does not directly use the sampling period data to update the associated edge weights, but instead retains the data as a sample to be confirmed and verifies it in subsequent sampling periods. The advantage of this embodiment is that abnormal sampling values ​​or state recording errors are not likely to directly affect the generation of the target collaborative control strategy, and the input data of the graph model remains traceable.

[0039] In a preferred embodiment, the generation of candidate cooperative control strategies further includes control strategy boundary compression processing. Instead of exhaustively listing all device action combinations in each control cycle, the control platform determines the candidate range based on the relationship between the node to be adjusted and its adjacent associated edges. For reactive power compensation switching sequences, only switching actions consistent with the current reactive power deviation direction are generated, and actions that would reverse the power factor offset direction are eliminated. For temperature control start-stop sequences, only actions that satisfy the temperature control start-stop holding constraints and match the direction of the thermal inertia mapping term are generated. For loop operation constraint adjustment sequences, only actions that will not cause deviations in current-carrying constraints are generated. The control platform combines the sequences after boundary compression processing to form a candidate cooperative control strategy set. If the candidate set is empty, the current device control state is maintained, and only the electrical-thermal coupling state diagram and action memory table are updated. If the candidate set is not empty, the calculation of control action propagation and constraint verification are performed. This processing method reduces the number of actions unrelated to the current energy efficiency degradation cause entering the evaluation process, avoiding an excessively large candidate space that would cause the control platform to process too many invalid combinations in a single control cycle. The advantage of this embodiment is that the candidate strategy generation is associated with the loss contribution node, the action propagation direction and the running constraints, and the control decision process has a stable data boundary.

[0040] Furthermore, in one embodiment, before issuing the target collaborative control strategy, control commands are encapsulated for consistency. The control platform converts reactive power compensation switching, temperature control start / stop, and loop operation constraint adjustment into control commands recognizable by the corresponding devices, and writes a control cycle identifier, action object identifier, action direction identifier, and constraint verification identifier for each control command. The control cycle identifier is used to avoid repeated execution of commands across cycles; the action object identifier is used to identify the device or loop corresponding to the control command; the action direction identifier is used to distinguish the direction of activation, deactivation, start, stop, or constraint adjustment; and the constraint verification identifier is used to record the pass status of voltage constraints, current-carrying constraints, temperature constraints, and action interval constraints. After issuing the control command, the control platform listens to the device feedback status. If the feedback status is consistent with the action direction identifier, the control command is marked as completed; if the feedback status does not return within the specified control cycle, the control command is marked as unconfirmed, and continuous reverse actions based on unconfirmed commands are prohibited. For unconfirmed commands, the control platform still collects relevant operating data in the next control cycle and determines whether a candidate collaborative control strategy needs to be regenerated based on changes in the operating data. The advantage of this embodiment is that when the control strategy is converted from the calculation result into the device control command, the constraint verification and cycle identifier are retained, which can reduce cross-cycle repetitive actions and continuous reverse actions.

[0041] In this embodiment, the electrical-thermal coupling state diagram, loss contribution, control action propagation, candidate collaborative control strategies, constraint verification sequences, and action memory table together constitute a complete closed loop for multi-equipment collaborative energy efficiency optimization in a prefabricated substation. After the operating data enters the control platform, a state vector is first formed, and then the state nodes and associated edges are updated. The control platform calculates the loss contribution of each state node based on the weight of the associated edges, filters the nodes to be adjusted, and generates candidate collaborative control strategies. After the candidate strategies undergo control action propagation calculation, conflict propagation direction identification, action vector merging, and constraint verification, a target collaborative control strategy is formed. After the target collaborative control strategy is issued and executed, the execution result is written into the action memory table and participates in the action cost calculation of the next control cycle. This closed-loop process revolves around the overall station loss, while preserving the control boundaries of voltage, current carrying capacity, temperature, and action continuity. The advantage of this embodiment is that the energy efficiency control of multiple devices in the prefabricated substation is transformed from decentralized threshold judgment to collaborative strategy screening under graph model constraints, which can achieve the unification of the station's energy efficiency control logic without changing the hardware structure.

Claims

1. A method for optimizing the energy efficiency of a prefabricated substation, characterized in that, include: Obtain the operating data of transformers, incoming and outgoing circuits, reactive power compensation equipment, station power loads and temperature control equipment in the prefabricated substation, and construct an electrical-thermal coupling state diagram based on the operating data to characterize the power transmission relationship, reactive power flow relationship, heat transfer relationship and auxiliary power consumption relationship; Calculate the loss contribution and control action propagation of each device based on the electrical-thermal coupling state diagram. Generate candidate coordinated control strategies that include reactive power compensation switching, temperature control start-stop, and loop operation constraint adjustment; A method for determining and executing a target collaborative control strategy with the minimum overall loss from the candidate collaborative control strategies, under the conditions of satisfying voltage, current carrying capacity, temperature, and action interval constraints.

2. The energy efficiency optimization method for a prefabricated substation according to claim 1, characterized in that, The construction of the electrical-thermal coupling state diagram includes mapping the transformer, the incoming and outgoing line circuits, the reactive power compensation equipment, the station power load, and the temperature control equipment as state nodes respectively. The active power transmission relationship between the transformer and the incoming and outgoing line circuits, the reactive power adjustment relationship between the reactive power compensation device and the incoming and outgoing line circuits, the heat load transfer relationship between the transformer and the temperature control device, and the auxiliary power consumption relationship between the temperature control device and the station power load are respectively mapped as associated edges; The weights of the associated edges are updated based on the voltage, current, active power, reactive power, load rate, and temperature data within the same sampling period.

3. The energy efficiency optimization method for a prefabricated substation according to claim 2, characterized in that, The calculation of the loss contribution includes: according to the state node type in the electrical-thermal coupling state diagram, the operating data is divided into transformer load loss data, circuit transmission loss data, reactive power flow loss data, station power consumption data, and temperature control power consumption data; Normalize and map the transformer load loss data, the circuit transmission loss data, the reactive power flow loss data, the station power consumption data, and the temperature control power consumption data within the same control cycle; Based on the weights of the associated edges, the normalized data is coupled and superimposed to obtain the loss contribution of each state node.

4. The energy efficiency optimization method for a prefabricated substation according to claim 1, characterized in that, The calculation of the control action propagation includes: establishing action vectors for the reactive power compensation switching, the temperature control start / stop, and the loop operation constraint adjustment, respectively; The motion vector is input into the electrical-thermal coupling state diagram, and the calculation is performed layer by layer along the reactive power flow relationship, the heat transfer relationship, and the auxiliary power consumption relationship. The voltage change, reactive power change, temperature change, and station power consumption change after propagation are jointly encoded to obtain the control action propagation quantity of the corresponding candidate collaborative control strategy; In this context, action vectors with conflict propagation directions within the same candidate collaborative control strategy are merged into a single action combination.

5. The energy efficiency optimization method for a prefabricated substation according to claim 1, characterized in that, The generation of the candidate collaborative control strategy includes: selecting state nodes whose loss contribution meets the preset contribution conditions as nodes to be adjusted, based on the loss contribution amount in the current control cycle as the sorting criteria. The set of cooperative devices is determined based on the adjacent associated edges of the node to be adjusted in the electrical-thermal coupling state diagram; Within the set of collaborative devices, generate reactive power compensation switching sequences, temperature control start / stop sequences, and loop operation constraint adjustment sequences; The reactive power compensation switching sequence, the temperature control start-stop sequence, and the loop operation constraint adjustment sequence are combined to form the candidate collaborative control strategy, and combinations that do not meet the equipment action interval constraints are eliminated.

6. The energy efficiency optimization method for a prefabricated substation according to claim 2, characterized in that, The weight update of the associated edge includes: writing the changes in active power, reactive power and temperature in adjacent sampling periods into the dynamic weight items of the corresponding associated edge. Write the transformer load rate, the switching status of the reactive power compensation equipment, and the start / stop status of the temperature control equipment into the state weight item of the corresponding associated edge; A weight correction factor is generated according to the deviation direction between the dynamic weight term and the state weight term of the same associated edge; The weights of the electrical energy transmission relationship, the reactive power flow relationship, the heat transfer relationship, and the auxiliary power consumption relationship are updated synchronously using the weight correction factor.

7. The energy efficiency optimization method for a prefabricated substation according to claim 3, characterized in that, The coupling superposition includes: establishing a nonlinear loss mapping term corresponding to the load rate for the transformer load loss data; A reactive power deviation mapping term corresponding to the power factor offset direction is established for the aforementioned reactive power flow loss data; A thermal inertia mapping term corresponding to the temperature change hysteresis is established for the aforementioned temperature control power consumption data; The nonlinear loss mapping term, the reactive power deviation mapping term, and the thermal inertia mapping term are multiplied by the weights of their respective associated edges and then input into the same loss contribution matrix. The loss contribution of each state node is then obtained based on the loss contribution matrix.

8. The energy efficiency optimization method for a prefabricated substation according to claim 4, characterized in that, The merging of action vectors with conflict propagation direction includes: identifying a first action vector in the same candidate cooperative control strategy that reduces reactive power change and increases station power consumption change, and a second action vector that reduces temperature change and increases station power consumption change. Generate an action priority order based on the loss contribution, remaining action interval time, and propagation level corresponding to the first action vector and the second action vector; Based on the action priority, the dominant action vector is retained, and the non-dominant action vectors are converted into delayed action vectors, forming a single action combination with a time sequence.

9. The energy efficiency optimization method for a prefabricated substation according to claim 2, characterized in that, After the weights of the associated edges are updated synchronously, the method further includes: obtaining the target collaborative control strategy of the previous control cycle and the corresponding execution data, and writing the difference in loss contribution before and after execution, the difference in control action propagation, and the difference in associated edge weights into the action memory table. When generating candidate collaborative control strategies for the current control cycle, the action cost is increased based on the action memory table for action combinations that are repeated and whose difference in loss contribution does not meet the convergence condition. The action cost is incorporated into the overall loss calculation process to obtain an evaluation value used to screen target collaborative control strategies.

10. The energy efficiency optimization method for a prefabricated substation according to claim 8, characterized in that, After forming the single action combination with time sequence, the method further includes: inputting the single action combination into a constraint verification sequence consisting of voltage constraint, current constraint, temperature constraint, reactive power compensation action interval constraint and temperature control start-stop hold constraint. When the single action combination fails the constraint verification sequence, the delayed action vector is replaced or the non-dominant action vector is deleted in sequence according to the action priority, and a new candidate cooperative control strategy is formed. When there are multiple candidate cooperative control strategies that pass the constraint verification sequence, the candidate cooperative control strategy with the smallest evaluation value is selected as the target cooperative control strategy.

Images