Photovoltaic inverter multi-target collaborative optimization control system and method

Abstract

The invention discloses a multi-objective collaborative optimization control system and method for a photovoltaic inverter, and relates to the technical field of inverter control. According to the method, multi-source state data of an inverter cluster and a power station environment are acquired, a cluster mirror image three-dimensional model is established according to the multi-source state data, state particles and compensation particles are set for each inverter, and according to historical operation records of the inverters, the state particles and the compensation particles are associated with a plurality of feature parameter ranges, so that the cluster mirror image three-dimensional model is established. State particles and compensation particles of different inverters are mutually connected and matched to obtain a state-compensation association network, real-time multi-source state data of an inverter cluster and a power station environment are obtained and input into the cluster mirror image three-dimensional model and the state-compensation association network, and then a multi-target collaborative decision is output. And simulating and executing the cluster mirror image three-dimensional model in the cluster mirror image three-dimensional model, and correcting the multi-target collaborative decision according to a simulation result.

Description

Technical Field

[0001] This invention relates to the field of inverter control technology, specifically to a multi-objective collaborative optimization control system and method for photovoltaic inverters. Background Technology

[0002] With the increasing global demand for clean energy, photovoltaic (PV) power generation has been widely adopted as a sustainable energy solution. As the core equipment in a PV power generation system, the PV inverter's main function is to convert the direct current (DC) generated by solar panels into alternating current (AC) for transmission to the grid or local loads. However, in the actual operation of PV power plants, inverter clusters face numerous challenges.

[0003] On the one hand, the operating status of inverters is affected by a variety of factors, including electrical parameters (such as voltage, current, and power), thermal performance (such as temperature and heat dissipation), and power plant environmental factors (such as light intensity and ambient temperature). These factors are interrelated and dynamically change, making it difficult to accurately monitor and assess the operating status of inverters. Traditional monitoring methods often only obtain a single or limited number of parameters, which cannot comprehensively reflect the actual operating status of the inverter and make it difficult to detect potential faults in advance.

[0004] On the other hand, how to quickly and effectively adjust and repair inverters when they malfunction or operate abnormally, in order to ensure the stable operation and efficient power generation of the entire inverter cluster, is an urgent problem to be solved. Traditional control methods usually control individual inverters independently, lacking coordinated optimization control of the inverter cluster, failing to fully utilize the overall advantages of the cluster, and proving inadequate in dealing with complex and changing operating conditions. Furthermore, fault propagation may occur between different inverters. If the fault propagation path cannot be identified and blocked in time, the fault range may expand, further affecting the power generation efficiency and reliability of the entire photovoltaic power plant. Therefore, this paper proposes a multi-objective coordinated optimization control system and method for photovoltaic inverters. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-objective collaborative optimization control system and method for photovoltaic inverters to address the shortcomings in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A multi-objective collaborative optimization control method for a photovoltaic inverter includes the following steps: Step S1: Obtain multi-source status data of the inverter cluster and the power station environment, and establish a cluster mirror 3D model based on the multi-source status data; Step S2: Set state particles and compensation particles for each inverter, and associate multiple characteristic parameter ranges with the state particles and compensation particles according to the inverter's historical operation records. Connect and match the state particles and compensation particles of different inverters to obtain the state-compensation correlation network. Step S3: Obtain real-time multi-source state data of the inverter cluster and power plant environment, and input them into the cluster mirror 3D model and state-compensation correlation network, and then output multi-objective collaborative decision. Simulate the execution of the cluster mirror 3D model in the cluster mirror 3D model, and correct the multi-objective collaborative decision based on the simulation results.

[0007] Furthermore, the process of acquiring multi-source state data includes: Multiple sensors are installed at key nodes of the inverter cluster and in the power plant environment. The key monitoring locations of the inverter cluster include the DC side, AC output side, heat sink, and inside the chassis of each inverter. The types of sensors used in inverter clusters include electrical sensors, thermal sensors, high-frequency response sensors, and image acquisition devices. The types of sensors used in the power plant environment include irradiators and ambient temperature sensors; The same sampling clock is set for all sensors, which is triggered in the form of a countdown. Then, whenever the sampling clock is reset to zero, the status data collected by each sensor is time-aligned, thereby obtaining multi-source status data of the inverter cluster and the power plant environment.

[0008] Furthermore, the process of establishing the cluster mirror 3D model includes: A cluster mirror 3D model is established based on the inverter cluster and the multi-source state data of the inverter cluster. The cluster mirror 3D model includes a physical structure region and an internal state region. The process of establishing the physical structure region includes: extracting the sampling data of the image acquisition unit from the multi-source state data, and then establishing a three-dimensional appearance contour model of each inverter based on the sampling data of the image acquisition unit. At the same time, based on the inverter model and known physical parameters, a three-dimensional model of its main internal components is constructed and embedded inside the three-dimensional appearance contour model. Meanwhile, the multi-source state data of the power plant environment is distributed in the form of point clouds inside and outside the three-dimensional appearance contour model. Each point cloud contains both the light intensity data collected by the irradiator and the temperature data collected by the ambient temperature sensor. The process of establishing the internal state region includes: The system retrieves multi-source state data from the inverter cluster, excluding the sampling data from the image acquisition unit. It then overlays the sampling data from the thermal sensor onto the corresponding 3D model inside the 3D appearance contour model with thermal patterns, and simultaneously maps the sampling data from the electrical sensor and the high-frequency response sensor onto the 3D model of the corresponding inverter to obtain the internal state region. The three-dimensional appearance contour models corresponding to each inverter are connected and laid out according to the actual electrical connection topology and geographical location of the inverter cluster. Based on the data contained in the external point cloud of the three-dimensional appearance contour model, the dark and temperature difference positions are marked on the three-dimensional appearance contour model, thus obtaining the cluster mirror three-dimensional model.

[0009] Furthermore, the process of setting the state particles includes: The state particles include high-efficiency running particles and multiple state-abnormal particles; The high-efficiency operating particle indicates that the inverter is currently in an ideal operating state; The abnormal state particles indicate that the inverter is currently experiencing one or more faults, and the fault names are labeled. Set a set of characteristic parameter ranges for the state particles, and obtain several historical operation records throughout the entire life cycle based on the inverter model in the inverter cluster, including historical operation records under normal operation and under one or more faults. For historical operation records under normal operating conditions, retrieve the same state data from the historical operation records within the same time interval, map them to the same two-dimensional coordinate system, perform normal distribution on the state data, and select the normal characteristic parameter range based on the normal distribution result; Set state collapse thresholds for various state data, construct normal change curves using the midpoint of the normal characteristic parameter range, differentiate the normal change curves for each time interval, and mark the time nodes in the derivative curves whose values ​​are greater than or equal to the state collapse threshold as state collapse points; otherwise, do not mark them. Simultaneously, based on historical operation records under one or more faults, the range of multiple abnormal characteristic parameters corresponding to various faults is obtained under different time intervals.

[0010] Furthermore, the process of setting the compensating particles includes: The compensation particles include power elastic particles, reactive power support particles, and harmonic compensation particles, which are used to output various inverter regulation decisions. The power elastic particle includes the ability to dynamically output multiple active power adjustment decisions based on the current active power adjustment range available from the inverter. The reactive power support particle includes dynamically outputting multiple reactive power adjustment decisions based on the reactive power adjustment range currently available from the inverter. The harmonic compensation particles contain various adjustment decisions regarding the operation of the active filter, such as reducing the operating efficiency of the active filter.

[0011] Furthermore, the process of establishing the state-compensation correlation network includes: All state particles and compensation particles are bound to the corresponding three-dimensional models in the cluster mirror three-dimensional model, and all historical operation records are retrieved to simulate the operation of the cluster mirror three-dimensional model. Based on the simulation results, the correlation between state particles of different inverters and the triggering relationship between state particles and compensation particles are established. First, multiple simulation time segments are set according to the state collapse time point, and the normal characteristic parameter range corresponding to each simulation time segment is used as the initial state data to initialize the cluster mirror 3D model. Then, based on the abnormal value fragments in the historical operation records with one or more faults, the cluster mirror 3D model is simulated to run abnormally. During the abnormal simulation operation, the relationship between state particles and compensation particles is analyzed, and the compensation particles are mobilized to correct the cluster mirror 3D model, so that the cluster mirror 3D model changes from the current running state to the initial running state. The relationships between state particles and compensating particles include the fault propagation chain between state particles, the effect relationship tree between state particles and compensating particles, and the synergistic antagonistic relationship chain between compensating particles. The fault propagation chain between state particles obtains the probability of occurrence between each state particle by analyzing and statistically analyzing different occurrence sequences, and sets a fault propagation chain between the state abnormal particles in the corresponding two three-dimensional models on the cluster mirror three-dimensional model, and records the occurrence probability as the propagation weight on the fault propagation chain. The process of establishing the effect relationship tree between state particles and compensating particles includes: during the abnormal simulation process, if a certain state particle is activated in an abnormal numerical segment, and a compensating particle is subsequently activated and produces a positive effect, then an effect relationship line is set between the corresponding state particle and the compensating particle. Then, after all the abnormal simulations are completed, the fault propagation chain between abnormal state particles and the effect relationship line between state particles and compensating particles are integrated to obtain the effect relationship tree.

[0012] Furthermore, the process of establishing the synergistic antagonistic relationship chain between compensating particles includes: Based on the connection relationships between state particles and compensating particles in the effect relationship tree, compensating particles connected to the same state particle are retrieved. Two or more compensating particles are retrieved sequentially for synchronous execution, and the positive effects of synchronous execution are compared with those of execution by a single compensating particle. If the positive effects of synchronous execution are better than those of execution by a single compensating particle, a cooperative connection line is set between the corresponding state particle and compensating particle in the effect relationship tree. If the positive effects of synchronous execution are the same as those of execution by a single compensating particle, no connection line is set. If the positive effects of synchronous execution are worse than those of execution by a single compensating particle, an antagonistic connection line is set between the corresponding state particle and compensating particle in the effect relationship tree, thus obtaining the state-compensation relationship network.

[0013] Furthermore, the process of inputting real-time multi-source state data into the cluster mirror 3D model and the state-compensation association network includes: Acquire real-time multi-source status data of inverter clusters and power plant environment, and dynamically update the cluster mirror 3D model based on the real-time multi-source status data; First, compare the real-time multi-source status data of each inverter with the high-efficiency operation particle. If it is determined that the real-time multi-source status data are all within the corresponding normal characteristic parameter range, then the corresponding inverter is determined to be operating normally. If it is determined that one or more real-time multi-source state data are not within the corresponding normal feature parameter range, then according to the type of real-time multi-source state data that is not within the corresponding normal feature parameter range, the corresponding state abnormal particle is matched, and then the abnormal feature parameter range contained in the state abnormal particle is compared with the real-time multi-source state data that is not within the corresponding normal feature parameter range. If real-time multi-source state data that is not within the corresponding normal feature parameter range completely conforms to the abnormal feature parameter range contained in a state abnormal particle, then the efficient running particle on the corresponding 3D model is replaced with the corresponding state abnormal particle, thereby activating the state-compensation correlation network.

[0014] Furthermore, the process of generating multi-objective collaborative decision-making includes: Based on the current total runtime of the corresponding inverter, the compensation particles associated with the abnormal state particles are retrieved from the state-compensation association network. At the same time, based on the cooperative / antagonistic connection lines between the compensation particles, the positive effects of one or more compensation particles being executed synchronously are obtained. Then, the combination of compensation particles with the best positive effects is selected to generate multi-objective cooperative decisions and execute them. Simultaneously, based on the propagation weights marked in the fault propagation chain of the abnormal state particles, a cumulative fault weight is set for the connected abnormal state particles, and a fault threshold is set. When the cumulative fault weight of an abnormal state particle is greater than or equal to the fault threshold, a multi-objective collaborative decision is output and executed based on the compensating particles connected to the abnormal state particles; otherwise, no operation is performed. Repeat the above process of generating and executing multi-objective collaborative decisions until the state-abnormal particles on the 3D model are replaced with efficient operating particles.

[0015] A multi-objective collaborative optimization control system for a photovoltaic inverter includes an inverter visualization module, an anomaly analysis module, and a multi-objective decision-making and control module; The inverter visualization module includes acquiring multi-source status data of the inverter cluster and the power plant environment, and establishing a cluster mirror 3D model based on the multi-source status data. The anomaly analysis module includes setting state particles and compensation particles for each inverter, associating multiple feature parameter ranges with the state particles and compensation particles based on the inverter's historical operation records, and connecting and matching the state particles and compensation particles of different inverters to obtain a state-compensation correlation network. The multi-objective decision-making and control module includes acquiring real-time multi-source state data of the inverter cluster and the power plant environment, inputting it into the cluster mirror 3D model and the state-compensation correlation network, and then outputting and executing multi-objective collaborative decisions. The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention enables refined management of the inverter's operating status by setting state particles and compensation particles and establishing a state-compensation correlation network. The state particles are associated with multiple characteristic parameter ranges, which can accurately identify the inverter's normal operating status and various fault states. At the same time, the fault propagation chain in the state-compensation correlation network can predict the propagation path and probability of faults in advance, realizing early warning of faults and helping maintenance personnel to take timely measures to avoid the expansion of the fault scope.

[0016] 2. This invention dynamically updates the cluster mirror 3D model based on real-time multi-source state data and generates multi-objective collaborative decision-making through a state-compensation correlation network. When an inverter malfunctions or operates abnormally, it can quickly select the optimal combination of compensating particles for adjustment, thereby achieving collaborative optimization control of the inverter cluster. This fully leverages the overall advantages of the inverter cluster, improves power generation efficiency, reduces operation and maintenance costs, and enhances the reliability and stability of the photovoltaic power station. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a flowchart of a multi-objective collaborative optimization control method for a photovoltaic inverter according to the present invention.

[0019] Figure 2 This is a system block diagram of a multi-objective collaborative optimization control system for a photovoltaic inverter according to the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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 embodiments of the present invention, not all embodiments. 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.

[0021] Please see Figure 1 As shown, a multi-objective collaborative optimization control method for a photovoltaic inverter includes the following steps: Step S1: Obtain multi-source status data of the inverter cluster and the power station environment, and establish a cluster mirror 3D model based on the multi-source status data; Step S2: Set state particles and compensation particles for each inverter, and associate multiple characteristic parameter ranges with the state particles and compensation particles according to the inverter's historical operation records. Connect and match the state particles and compensation particles of different inverters to obtain the state-compensation correlation network. Step S3: Obtain real-time multi-source state data of the inverter cluster and power plant environment, and input them into the cluster mirror 3D model and state-compensation correlation network, and then output multi-objective collaborative decision. Simulate the execution of the cluster mirror 3D model in the cluster mirror 3D model, and correct the multi-objective collaborative decision based on the simulation results.

[0022] Furthermore, step S1 is implemented through the following process: Step S101: Obtain multi-source state data. The specific process includes: Multiple sensors are installed at key nodes of the inverter cluster and in the power plant environment. The key monitoring locations of the inverter cluster include the DC side, AC output side, heat sink, and inside the chassis of each inverter. The types of sensors used in inverter clusters include electrical sensors, thermal sensors, high-frequency response sensors, and image acquisition devices. The types of sensors used in the power plant environment include irradiators and ambient temperature sensors; The same sampling clock is set for all sensors. The sampling clock is triggered in the form of a countdown, and the duration is generally set to 1 to 5 seconds. Then, when the sampling clock is reset to zero, the status data collected by each sensor is time-aligned, thereby obtaining multi-source status data of the inverter cluster and the power plant environment.

[0023] Step S102: Establish a 3D model of the cluster image. The specific process includes: A cluster mirror 3D model is established based on the inverter cluster and the multi-source state data of the inverter cluster. The cluster mirror 3D model includes a physical structure region and an internal state region. The process of establishing the physical structure region includes: extracting the sampling data of the image acquisition unit from the multi-source state data, and then establishing a three-dimensional appearance contour model of each inverter based on the sampling data of the image acquisition unit. At the same time, based on the inverter model and known physical parameters, constructing a three-dimensional model of its main internal components (such as capacitors, inductors, and power modules) and embedding it inside the three-dimensional appearance contour model. Meanwhile, the multi-source state data of the power plant environment is distributed in the form of point clouds inside and outside the three-dimensional appearance contour model. It should be noted that each point cloud contains both the light intensity data collected by the irradiation instrument and the temperature data collected by the ambient temperature sensor. The process of establishing the internal state region includes: The system retrieves multi-source state data from the inverter cluster, excluding the sampling data from the image acquisition unit. It then overlays the sampling data from the thermal sensor onto the corresponding 3D model inside the 3D appearance contour model with thermal patterns, and simultaneously maps the sampling data from the electrical sensor and the high-frequency response sensor onto the 3D model of the corresponding inverter to obtain the internal state region. The three-dimensional appearance contour models corresponding to each inverter are connected and laid out according to the actual electrical connection topology of the inverter cluster (such as series, parallel, combiner box, etc.) and geographical location. Based on the data contained in the external point cloud of the three-dimensional appearance contour model, the dark and temperature difference positions are marked on the three-dimensional appearance contour model, thus obtaining the cluster mirror three-dimensional model.

[0024] Furthermore, step S2 is implemented through the following process: Step S201: Set the state particles and compensation particles. The specific process includes: The state particles include high-efficiency running particles and multiple state-abnormal particles; The high-efficiency operating particle indicates that the inverter is currently in an ideal operating state; The abnormal state particles indicate that the inverter is currently experiencing one or more faults, and the fault names are labeled. A set of characteristic parameter ranges is set for the state particles, such as light intensity range, temperature rise range, amplitude change range, etc. The process of obtaining the characteristic parameter range includes: Based on the inverter model in the inverter cluster, obtain several historical operation records throughout the entire life cycle, including historical operation records under normal operation and under one or more faults. For historical operation records under normal operating conditions, retrieve the same state data from the historical operation records under the same time interval and map them to the same two-dimensional coordinate system. Perform normal distribution on the state data, and select the central position and the numerical intervals on the left and right sides that are 10% of the entire numerical interval based on the normal distribution results as the normal characteristic parameter range under the corresponding time interval. Set state collapse thresholds for various state data, construct normal change curves using the midpoint of the normal characteristic parameter range, differentiate the normal change curves for each time interval, and mark the time nodes in the derivative curves whose values ​​are greater than or equal to the state collapse threshold as state collapse points; otherwise, do not mark them. For historical operation records under one or more faults, the range of each normal characteristic parameter is compared with the status data in the historical operation record according to the time interval order. The numerical segments of each status data that are not in the corresponding normal characteristic parameter range are extracted and recorded as abnormal numerical segments. The number of times abnormal numerical segments of each status data occur is counted. Set a frequency threshold. If the number of times an abnormal value segment occurs is greater than or equal to the frequency threshold, then the corresponding type of status data will be marked as the characteristic abnormal status data of the corresponding fault; otherwise, no marking will be made. Then, the characteristic abnormal state data in each historical operation record are distributed normally, and based on the normal distribution results, the range of multiple abnormal characteristic parameters corresponding to various faults in different time intervals are obtained.

[0025] The compensation particles include power elastic particles, reactive power support particles, and harmonic compensation particles, which are used to output various inverter regulation decisions. The power elastic particle includes the ability to dynamically output multiple active power adjustment decisions based on the current active power adjustment range available from the inverter. The reactive power support particle includes dynamically outputting multiple reactive power adjustment decisions based on the reactive power adjustment range currently available from the inverter. The harmonic compensation particles contain various adjustment decisions regarding the operation of the active filter, such as reducing the operating efficiency of the active filter.

[0026] Step S202: Establish a state-compensation association network, the specific process of which includes: All state particles and compensation particles are bound to the corresponding three-dimensional models in the cluster mirror three-dimensional model, and all historical operation records are retrieved to simulate the operation of the cluster mirror three-dimensional model. Based on the simulation results, the correlation between state particles of different inverters and the triggering relationship between state particles and compensation particles are established. The simulation process includes: First, multiple simulation time segments are set according to the state collapse time point, and the normal characteristic parameter range corresponding to each simulation time segment is used as the initial state data to initialize the cluster mirror 3D model. Then, based on the abnormal value fragments in the historical operation records with one or more faults, the cluster mirror 3D model is simulated to run abnormally. During the abnormal simulation operation, the relationship between state particles and compensation particles is analyzed, and the compensation particles are mobilized to correct the cluster mirror 3D model, so that the cluster mirror 3D model changes from the current running state to the initial running state. The relationships between state particles and compensating particles include the fault propagation chain between state particles, the effect relationship tree between state particles and compensating particles, and the synergistic antagonistic relationship chain between compensating particles. The fault propagation chain between state particles is analyzed and statistically analyzed based on different occurrence sequences. For example, through analysis and statistics, it was found that when the overheating state abnormal particle of inverter A is activated, there is an 85% probability that it will lead to the activation of the overvoltage state abnormal particle of its downstream inverter B within 5 minutes. Therefore, a fault propagation chain is set between the state abnormal particles of the corresponding two 3D models on the cluster mirror 3D model, and the occurrence probability is recorded as the propagation weight and marked on the fault propagation chain. The process of establishing the effect relationship tree between state particles and compensating particles includes: during the abnormal simulation operation, if a certain state particle is activated in an abnormal numerical segment, and a compensating particle is subsequently activated and produces a positive effect, then an effect relationship line is set between the corresponding state particle and the compensating particle. For example, positive effects such as the total power recovery or temperature drop of the inverter are set between the corresponding state particle and the compensating particle. Then, after all the abnormal simulations are completed, the fault propagation chain between abnormal state particles and the effect relationship line between state particles and compensating particles are integrated to obtain the effect relationship tree. The process of establishing a synergistic antagonistic relationship chain between compensating particles includes: Based on the connection relationships between state particles and compensating particles in the effect relationship tree, compensating particles connected to the same state particle are retrieved. Two or more compensating particles are retrieved sequentially for synchronous execution, and the positive effects of synchronous execution are compared with those of execution by a single compensating particle. If the positive effects of synchronous execution are better than those of execution by a single compensating particle, a cooperative connection line is set between the corresponding state particle and compensating particle in the effect relationship tree. If the positive effects of synchronous execution are the same as those of execution by a single compensating particle, no connection line is set. If the positive effects of synchronous execution are worse than those of execution by a single compensating particle, an antagonistic connection line is set between the corresponding state particle and compensating particle in the effect relationship tree, thus obtaining the state-compensation relationship network.

[0027] Furthermore, step S3 is implemented through the following process: Step S301: Input real-time multi-source state data into the cluster mirror 3D model and the state-compensation association network. The specific process includes: Acquire real-time multi-source status data of inverter clusters and power plant environment, and dynamically update the cluster mirror 3D model based on the real-time multi-source status data; First, compare the real-time multi-source status data of each inverter with the high-efficiency operation particle. If it is determined that the real-time multi-source status data are all within the corresponding normal characteristic parameter range, then the corresponding inverter is determined to be operating normally. If it is determined that one or more real-time multi-source state data are not within the corresponding normal feature parameter range, then according to the type of real-time multi-source state data that is not within the corresponding normal feature parameter range, the corresponding state abnormal particle is matched, and then the abnormal feature parameter range contained in the state abnormal particle is compared with the real-time multi-source state data that is not within the corresponding normal feature parameter range. If real-time multi-source state data that is not within the corresponding normal feature parameter range completely conforms to the abnormal feature parameter range contained in a state abnormal particle, then the efficient running particle on the corresponding 3D model is replaced with the corresponding state abnormal particle, thereby activating the state-compensation correlation network.

[0028] Step S302: Generate multi-objective collaborative decision-making, the specific process of which includes: Based on the current total runtime of the corresponding inverter, the compensation particles associated with the abnormal state particles are retrieved from the state-compensation association network. At the same time, based on the cooperative / antagonistic connection lines between the compensation particles, the positive effects of one or more compensation particles being executed synchronously are obtained. Then, the combination of compensation particles with the best positive effects is selected to generate multi-objective cooperative decisions and execute them. Simultaneously, based on the propagation weights marked in the fault propagation chain of the abnormal state particles, a cumulative fault weight is set for the connected abnormal state particles, and a fault threshold is set. When the cumulative fault weight of an abnormal state particle is greater than or equal to the fault threshold, a multi-objective collaborative decision is output and executed based on the compensating particles connected to the abnormal state particles; otherwise, no operation is performed. Repeat the above process of generating and executing multi-objective collaborative decisions until the state-abnormal particles on the 3D model are replaced with efficient operating particles.

[0029] Please see Figure 2 As shown, a multi-objective collaborative optimization control system for a photovoltaic inverter includes an inverter visualization module, an anomaly analysis module, and a multi-objective decision-making and control module. The inverter visualization module includes acquiring multi-source status data of the inverter cluster and the power plant environment, and establishing a cluster mirror 3D model based on the multi-source status data. The anomaly analysis module includes setting state particles and compensation particles for each inverter, associating multiple feature parameter ranges with the state particles and compensation particles based on the inverter's historical operation records, and connecting and matching the state particles and compensation particles of different inverters to obtain a state-compensation correlation network. The multi-objective decision control module includes acquiring real-time multi-source state data of the inverter cluster and the power plant environment, inputting it into the cluster mirror 3D model and the state-compensation correlation network, and then outputting and executing multi-objective collaborative decisions.

[0030] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for multi-objective collaborative optimization control of a photovoltaic inverter, characterized in that, The method comprises the following steps: Step S1, acquiring multi-source state data of the inverter cluster and the power plant environment, and establishing a cluster mirror three-dimensional model according to the multi-source state data; Step S2, setting state particles and compensation particles for each inverter, and associating a plurality of characteristic parameter ranges with the state particles and the compensation particles according to historical operation records of the inverter, and connecting and matching the state particles and the compensation particles of different inverters to obtain a state-compensation association network; Step S3, acquiring real-time multi-source state data of the inverter cluster and the power plant environment, and inputting the real-time multi-source state data into the cluster mirror three-dimensional model and the state-compensation association network, and then outputting a multi-target collaborative decision, simulating execution of the cluster mirror three-dimensional model in the cluster mirror three-dimensional model, and correcting the multi-target collaborative decision according to a simulation result.

2. The photovoltaic inverter multi-objective collaborative optimization control method according to claim 1, characterized in that, The acquisition process of the multi-source state data comprises: A plurality of sensors are arranged at key nodes of the inverter cluster and in the power plant environment, and the sensors arranged at the inverter cluster include electrical sensors, thermal sensors, high-frequency response sensors and image acquisition devices; The same sampling clock is arranged for all the sensors, wherein the sampling clock is triggered in the form of countdown, and then when the sampling clock resets at zero, the state data collected by each sensor is time-aligned, and then the multi-source state data of the inverter cluster and the power plant environment is obtained.

3. The photovoltaic inverter multi-objective collaborative optimization control method according to claim 2, characterized in that, The establishment process of the cluster mirror three-dimensional model comprises: The cluster mirror three-dimensional model is established according to the inverter cluster and the multi-source state data of the inverter cluster, and the cluster mirror three-dimensional model comprises a physical structure region and an internal state region; The sampling data of the image acquisition unit is extracted from the multi-source state data, and then a three-dimensional appearance contour model of each inverter is established according to the sampling data of the image acquisition unit, and meanwhile a three-dimensional model of the main internal components of the inverter is constructed according to the model and known physical parameters of the inverter, and is embedded in the three-dimensional appearance contour model, and meanwhile the multi-source state data of the power plant environment is distributed in the form of point cloud inside and outside the three-dimensional appearance contour model; The multi-source state data of the inverter cluster except the sampling data of the image acquisition unit is called, the sampling data of the thermal sensor is overlaid on the corresponding three-dimensional model inside the three-dimensional appearance contour model in the form of thermal lines, and the sampling data of the electrical sensor and the high-frequency response sensor is simultaneously mapped to the three-dimensional model of the corresponding inverter, and then the internal state region is obtained; The three-dimensional appearance contour models of the inverters are connected and laid out according to the actual electrical connection topological relationship and geographical position relationship of the inverter cluster.

4. The photovoltaic inverter multi-objective collaborative optimization control method according to claim 3, characterized in that, The setting process of the state particles comprises: The state particles comprise high-efficiency operation particles and a plurality of state abnormal particles; The high-efficiency operation particles represent that the current inverter is in an ideal operation state; The state abnormal particles represent that the current inverter is in one or more faults, and the fault name is marked; A group of characteristic parameter ranges are set for the state particles, and a plurality of historical operation records in normal operation states and in one or more faults are acquired according to the inverter models in the inverter cluster; For the historical operation record in normal operation state, the same state data in the historical operation record in the same time interval is mapped in the same two-dimensional coordinate system, the state data is normally distributed, and the normal characteristic parameter range is selected according to the normal distribution result; The state collapse threshold is set for various state data, the middle value of the normal characteristic parameter range is used to construct the normal change curve, the derivative of the normal change curve in each time interval is calculated, the time node corresponding to the value in the derivative result curve greater than or equal to the state collapse threshold is marked as the state collapse point, otherwise, no marking is performed; Meanwhile, according to the historical operation record under one or more faults, the abnormal characteristic parameter range corresponding to various faults in different time intervals is obtained.

5. The photovoltaic inverter multi-objective collaborative optimization control method according to claim 4, characterized in that, The setting process of the compensating particle includes: The compensating particle includes a power elastic particle, a reactive power support particle and a harmonic compensation particle, which are used to output various inverter adjustment decisions; The power elastic particle contains a plurality of active power adjustment decisions dynamically output according to the current active power adjustment range that can be provided by the inverter; The reactive power support particle contains a plurality of reactive power adjustment decisions dynamically output according to the current reactive power adjustment range that can be provided by the inverter; The harmonic compensation particle contains a plurality of adjustment decisions about the operation of the active filter, such as reducing the operation efficiency of the active filter.

6. The photovoltaic inverter multi-objective collaborative optimization control method according to claim 5, characterized in that, The establishment process of the state-compensation association network includes: According to the state collapse point, a plurality of simulated time sections are set, and the normal characteristic parameter range corresponding to each simulated time section is used as initial state data to perform initial operation state on the cluster mirror three-dimensional model, and then the cluster mirror three-dimensional model is simulated and operated abnormally according to the abnormal value segment in the historical operation record with one or more faults; The association relationship between the state particle and the compensating particle includes a fault propagation chain between the state particles, an effect relationship tree between the state particles and the compensating particles, and a synergistic antagonistic relationship chain between the compensating particles; The fault propagation chain between the state particles is obtained by analyzing and counting the appearance probability between the state particles in different appearance sequences, and the fault propagation chain is set between the state abnormal particles on the corresponding two three-dimensional models on the cluster mirror three-dimensional model, and the appearance probability is marked as a propagation weight on the fault propagation chain; For the effect relationship tree between the state particles and the compensating particles, if it is judged that a state particle is activated in an abnormal value segment, and a compensating particle is subsequently activated and produces a positive effect during the abnormal simulation operation, an effect relationship line is set between the corresponding state particle and the compensating particle; Then, after all abnormal simulation operations are completed, the fault propagation chain between the state abnormal particles and the effect relationship line between the state particles and the compensating particles are integrated to obtain the effect relationship tree.

7. The photovoltaic inverter multi-objective collaborative optimization control method according to claim 6, characterized in that, The establishment process of the synergistic antagonistic relationship chain between the compensating particles includes: According to the connection relationship between the state particles and the compensation particles in the effect relationship tree, the compensation particles connected with the same state particle are called, two or more compensation particles are called in turn for synchronous execution, and the positive effects of the synchronous execution and the single compensation particle execution are compared. If it is judged that the positive effect of the synchronous execution is better than that of the single compensation particle execution, a synergistic connection line is set between the corresponding state particle and the compensation particle in the effect relationship tree. If it is judged that the positive effect of the synchronous execution is consistent with that of the single compensation particle execution, no connection line is set. If it is judged that the positive effect of the synchronous execution is worse than that of the single compensation particle execution, an antagonistic connection line is set between the corresponding state particle and the compensation particle in the effect relationship tree, and then the state-compensation association network is obtained.

8. The photovoltaic inverter multi-objective collaborative optimization control method according to claim 7, characterized in that, The process of inputting real-time multi-source state data into the cluster mirror three-dimensional model and the state-compensation association network includes: Obtain real-time multi-source state data of the inverter cluster and the power plant environment, and dynamically update the cluster mirror three-dimensional model according to the real-time multi-source state data; First, compare the real-time multi-source state data of each inverter with the high-efficiency running particle. If it is judged that the real-time multi-source state data is within the corresponding normal characteristic parameter range, it is judged that the corresponding inverter is running normally. If it is judged that one or more real-time multi-source state data is not within the corresponding normal characteristic parameter range, match the corresponding state abnormal particle according to the type of the real-time multi-source state data not within the corresponding normal characteristic parameter range, and then compare the abnormal characteristic parameter range contained in the state abnormal particle with the real-time multi-source state data not within the corresponding normal characteristic parameter range. If the real-time multi-source state data not within the corresponding normal characteristic parameter range completely meets the abnormal characteristic parameter range contained in the state abnormal particle, replace the high-efficiency running particle on the corresponding three-dimensional model with the corresponding state abnormal particle, and then start the state-compensation association network.

9. The photovoltaic inverter multi-objective collaborative optimization control method according to claim 8, characterized in that, The generation process of the multi-target collaborative decision includes: According to the current total running time of the corresponding inverter, the compensation particles associated with the state abnormal particles are called from the state-compensation association network, and the positive effects under the synchronous execution of single or multiple compensation particles are obtained according to the synergistic / antagonistic connection lines between the compensation particles, and then the compensation particle combination with the best positive effect is selected to generate a multi-target collaborative decision and execute it. At the same time, according to the propagation weight value marked by the fault propagation chain of the state abnormal particle, the cumulative fault weight value of the connected state abnormal particle is set, and the fault threshold is set. When the cumulative fault weight value of the state abnormal particle is greater than or equal to the fault threshold, the multi-target collaborative decision is output and executed according to the compensation particles connected with the state abnormal particle, otherwise no operation is performed. Repeat the process of generating and executing the multi-target collaborative decision until the state abnormal particle on the three-dimensional model is replaced by the high-efficiency running particle.

10. A photovoltaic inverter multi-objective collaborative optimization control system for implementing the photovoltaic inverter multi-objective collaborative optimization control method of any one of claims 1-9, characterized in that, It includes an inverter visualization module, an abnormal analysis module, and a multi-target decision control module. The inverter visualization module includes obtaining multi-source state data of the inverter cluster and the power station environment, and establishing a cluster mirror three-dimensional model according to the multi-source state data. The abnormality analysis module comprises setting state particles and compensation particles for each inverter, and associating a plurality of characteristic parameter ranges with the state particles and the compensation particles according to historical operation records of the inverter, and connecting and matching the state particles and the compensation particles of different inverters to obtain a state-compensation association network; The multi-target decision regulation module comprises acquiring real-time multi-source state data of the inverter cluster and the power station environment, and inputting the real-time multi-source state data into the cluster mirror three-dimensional model and the state-compensation association network, and then outputting a multi-target cooperative decision and executing the multi-target cooperative decision.

Images