CN-122000974-A - Power distribution method and system of energy storage equipment

CN122000974ACN 122000974 ACN122000974 ACN 122000974ACN-122000974-A

Abstract

The invention provides a power distribution method and a system of energy storage equipment, wherein the method comprises the steps of identifying the current operation condition based on power grid frequency data, electricity price signal data and the system power demand change rate; the method comprises the steps of dynamically adjusting weight coefficients corresponding to all optimization targets according to recognized operation conditions to obtain working condition self-adaptive weights, inputting health state data, temperature data and cycle number data of all energy storage units into a preset nonlinear health mapping network to obtain health state estimated values and high-dimensional implicit state vectors of all the energy storage units, inputting index type features, imaging features, working condition self-adaptive weights and corrected health state estimated values into a multi-model integrated unit to obtain power distribution weights of all the energy storage units, and carrying out power distribution on all the energy storage units according to the power distribution weights. According to the invention, through the self-adaptive weight adjustment of working conditions and the dynamic switching of the optimization target, the power distribution strategy can adapt to different operation scenes such as power grid frequency fluctuation, electricity price change and the like.

Inventors

YU LIANGUANG
WANG CHANGYE
TIAN ZHENHUA

Assignees

济南朗瑞电气有限公司

Dates

Publication Date: 20260508
Application Date: 20260410

Claims (10)

1. A method of power distribution for an energy storage device, comprising the steps of: acquiring operation data, power grid frequency data and electricity price signal data of each energy storage unit in energy storage equipment; Extracting operation morphological characteristics, power fluctuation characteristics and operation correlation characteristics based on the operation data, constructing index type characteristics, and converting a typical operation curve into imaging characteristics; Identifying a current operating condition based on the grid frequency data, the electricity price signal data and the system power demand change rate; According to the identified operation conditions, dynamically adjusting weight coefficients corresponding to all optimization targets to obtain condition self-adaptive weights; Inputting the health state data, the temperature data and the cycle number data of each energy storage unit into a preset nonlinear health mapping network to obtain the health state estimated value and the high-dimensional implicit state vector of each energy storage unit; inputting the health state estimated value and the high-dimensional implicit state vector into a second long-short-term memory network to obtain the health state evolution rate of each energy storage unit, and constructing a composite loss function based on distribution error loss, evolution consistency loss and monotonicity constraint loss to correct the mapping result; inputting the index type characteristics, the imaging characteristics, the working condition self-adaptive weights and the corrected health state estimated values into a multi-model integrated unit to obtain the power distribution weights of the energy storage units; And carrying out power distribution on each energy storage unit according to the power distribution weight.
2. The method of claim 1, wherein extracting the extracted operational morphology features comprises: calculating average power, power peak-to-average ratio and charge-discharge depth ratio of the whole day according to a typical operation curve of the energy storage unit; Dividing a day into a plurality of time periods, and respectively calculating the average power of each time period in a typical running curve; calculating the maximum value and the minimum value of the average power of each period according to the historical operation data, and calculating the first difference ratio of the maximum value of the average power of each period to the average power of the period corresponding to the typical operation curve and the second difference ratio of the minimum value of the average power of each period to the average power of the period corresponding to the typical operation curve; And combining the whole-day average power, the power peak-to-average ratio, the charge-discharge depth ratio, the average power of each period, the first difference ratio and the second difference ratio into an operation morphological characteristic.
3. The method of claim 1, wherein extracting the power fluctuation feature comprises: Calculating the standard deviation of the instantaneous power in each time period, and averaging all running periods to obtain the daily fluctuation standard deviation of the time period; Calculating the power standard deviation of the same time in different operation periods for each period, and averaging all the times in the period to obtain a cross-period stability index of the period; and forming a power fluctuation characteristic by using the standard deviation of the intra-day fluctuation and the cross-period stability index of each period.
4. The method of claim 1, wherein extracting the operational correlation feature comprises: calculating a correlation coefficient between the output power of the energy storage unit and the frequency of the power grid to obtain a first correlation coefficient; Calculating a correlation coefficient of the output power of the energy storage unit and the electricity price signal to obtain a second phase relation number; separating working day and holiday operation data, respectively constructing a working day typical operation curve and a holiday typical operation curve, and calculating a correlation coefficient between the working day typical operation curve and the holiday typical operation curve to obtain a third phase relation number; And combining the first correlation coefficient, the second correlation coefficient and the third correlation coefficient into an operation correlation characteristic.
5. The method of claim 1, wherein converting the representative operating curve into an imaging signature comprises: Constructing an improved recursion diagram according to a typical running curve, wherein the improved recursion diagram is used as a red channel of a three-channel color image; Scaling the typical running curve to a preset interval, and generating an angle difference sine matrix serving as a green channel through a gram difference angle field algorithm; discretizing a typical operation curve into a plurality of fractional states, counting first-order transition probabilities among the states, and constructing a Markov transition field matrix serving as a blue channel; And combining to generate a three-channel color imaging feature based on the red channel, the green channel and the blue channel.
6. The method of claim 1, wherein the identifying the current operating condition comprises: Constructing an input feature vector based on the power grid frequency fluctuation, the electricity price signal and the system power demand change rate; processing the input feature vector by adopting a fuzzy reasoning system, and outputting the identification result of the current operation condition; The operation working condition comprises at least one of a peak clipping and valley filling mode, a demand response mode and an emergency standby mode.
7. The method of claim 1, wherein the nonlinear health mapping network comprises: The feature word segmentation device is used for distributing independent weight vectors and bias vectors for each input feature in the health state data, the temperature data and the cycle number data of each energy storage unit, and projecting data with different physical attributes to a unified high-dimensional semantic space to obtain feature embedding vectors; the deep learning module comprises a multi-head attention mechanism and a feedforward neural network and is used for carrying out global feature interaction processing on the feature embedded vector to obtain an updated feature vector; And the first long-term and short-term memory network is used for carrying out time sequence prediction processing on the updated feature vector to obtain the health state estimated value and the high-dimensional implicit state vector of each energy storage unit.
8. The method of claim 1, wherein constructing a composite loss function comprises: constructing distribution error loss for quantifying deviation between power distribution value and system demand; constructing evolution consistency loss for restraining the predicted value of the evolution rate of the health state to keep consistent with the actual evolution rate; constructing monotonicity constraint loss for punishing power allocation decisions violating monotonicity decreasing physical laws of health states; and carrying out weighted summation on the distribution error loss, the evolution consistency loss and the monotonicity constraint loss to obtain a composite loss function.
9. The method of claim 1, wherein the multi-model integrated unit comprises: The plurality of base learners at least comprise a linear mapping model based on health state, a nonlinear mapping model based on health state, a support vector machine model based on power fluctuation characteristics and a random forest model based on operation morphological characteristics; The element learner adopts a decision tree model and is used for carrying out integration and fusion on the output of each base learner; The method comprises the steps that each base learner acquires an out-of-refraction prediction result in a cross training mode, and the meta learner trains by taking the out-of-refraction prediction result of each base learner as input.
10. A power distribution system of an energy storage device, characterized in that it comprises a power distribution method of an energy storage device according to any of claims 1-9, comprising: the data acquisition module is used for acquiring operation data, power grid frequency data and electricity price signal data of each energy storage unit in the energy storage equipment; The characteristic construction module is used for extracting operation morphological characteristics, power fluctuation characteristics and operation correlation characteristics based on the operation data, constructing index type characteristics and converting a typical operation curve into imaging characteristics; The working condition identification module is used for identifying the current operation working condition based on the power grid frequency data, the electricity price signal data and the system power demand change rate; The dynamic weight adjustment module is used for dynamically adjusting the weight coefficient corresponding to each optimization target according to the identified operation condition to obtain a condition self-adaptive weight; The nonlinear health mapping module comprises a feature word segmentation device, a deep learning module and a first long-term and short-term memory network and is used for mapping health state data of each energy storage unit into a health state estimated value and a high-dimensional implicit state vector; the evolution rate constraint module comprises a second long-term and short-term memory network and is used for predicting the evolution rate of the health state and correcting the mapping result based on the composite loss function; the multi-model integrated module comprises a plurality of basic learners and meta learners and is used for determining the power distribution weight of each energy storage unit; and the power distribution execution module is used for distributing power to each energy storage unit according to the power distribution weight.

Description

Power distribution method and system of energy storage equipment Technical Field The invention relates to the technical field of energy management of electrochemical energy storage systems, in particular to a power distribution method and system of energy storage equipment. Background With the continuous expansion of renewable energy grid-connected scale and the increasing demand for flexibility of power systems, electrochemical energy storage technology has become a key infrastructure for supporting safe and stable operation of power grids. Currently, a proportional distribution method based on a state of charge or rated capacity is generally adopted for power distribution of an energy storage system, and a part of advanced systems introduce simple health state estimation as a correction coefficient. However, the existing feature extraction method is difficult to effectively describe the overall view of the operation characteristics of the energy storage unit, lacks systematic characterization on a power fluctuation mode, an operation morphology rule and external association factors, causes insufficient information basis of a power distribution decision, cannot adapt to multi-scene switching requirements such as peak clipping and valley filling, demand response, frequency support and the like, has response lag under working conditions such as power grid frequency out-of-limit or severe fluctuation of electricity price and the like, cannot be dynamically adjusted according to real-time working conditions such as power grid frequency, electricity price signals, power demand change and the like, is difficult to adapt to complex and changeable power grid operation scenes, has strong linear assumption of a traditional health state assessment model, is difficult to capture nonlinear influences of multi-factor coupling such as temperature, multiplying power and the like, does not consider time sequence evolution trend of a health state, easily causes assessment results to deviate from an actual aging state, further causes power distribution unbalance and accelerates the integral performance degradation of the energy storage system. Therefore, the invention designs a power distribution method and a power distribution system for energy storage equipment. Disclosure of Invention Therefore, the technical problem to be solved by the present invention is to overcome the above-mentioned drawbacks, and thus provide a power distribution method and system for an energy storage device. In order to solve the above-mentioned problems, the present invention provides a power distribution method of an energy storage device, which includes the following steps: acquiring operation data, power grid frequency data and electricity price signal data of each energy storage unit in energy storage equipment; Extracting operation morphological characteristics, power fluctuation characteristics and operation correlation characteristics based on the operation data, constructing index type characteristics, and converting a typical operation curve into imaging characteristics; Identifying a current operating condition based on the grid frequency data, the electricity price signal data and the system power demand change rate; According to the identified operation conditions, dynamically adjusting weight coefficients corresponding to all optimization targets to obtain condition self-adaptive weights; Inputting the health state data, the temperature data and the cycle number data of each energy storage unit into a preset nonlinear health mapping network to obtain the health state estimated value and the high-dimensional implicit state vector of each energy storage unit; inputting the health state estimated value and the high-dimensional implicit state vector into a second long-short-term memory network to obtain the health state evolution rate of each energy storage unit, and constructing a composite loss function based on distribution error loss, evolution consistency loss and monotonicity constraint loss to correct the mapping result; inputting the index type characteristics, the imaging characteristics, the working condition self-adaptive weights and the corrected health state estimated values into a multi-model integrated unit to obtain the power distribution weights of the energy storage units; And carrying out power distribution on each energy storage unit according to the power distribution weight. Preferably, the extracting the operation morphological feature includes: calculating average power, power peak-to-average ratio and charge-discharge depth ratio of the whole day according to a typical operation curve of the energy storage unit; Dividing a day into a plurality of time periods, and respectively calculating the average power of each time period in a typical running curve; calculating the maximum value and the minimum value of the average power of each period according to the historical operation data, and calculating the first dif