JP3852371B2 - Secondary battery charge rate estimation device - Google Patents

Description

（数１３）式において、モデル入力は電流Ｉ［Ａ］（正値は充電、負値は放電）、モデル出力は端子電圧Ｖ［Ｖ］、Ｒ_１〔Ω］は電荷移動抵抗、Ｒ_２［Ω］は純抵抗、Ｃ_１［Ｆ］は電気二重層容量、Ｖ_０［Ｖ］は開路電圧である。なお、ｓはラプラス演算子である。本モデルは、正極、負極を特に分離していないリダクションモデル（一次）であるが、実際の電池の充放電特性を比較的正確に示すことが可能である。このように本実施例においては、電池モデルの次数を１次にした構成を例として説明する。
【００１２】
上記（数１３）式の電池モデルから適応デジタルフィルタまでの導出を最初に説明する。
（数１３）式を変形すると（数１４）式（＝数７）になる。
【００１３】
【数１４】

ただし、Ｔ_１＝Ｃ_１・Ｒ_１、 Ｔ_２＝Ｃ_１・Ｒ_１・Ｒ_２／（Ｒ_１＋Ｒ_２）
Ｋ＝Ｒ_１＋Ｒ_２
開路電圧Ｖ_０は、電流Ｉに可変な効率ｄを乗じたものを、ある初期状態から積分したものと考えれば、（数１５）式（＝数２＝数８）で書ける。
【００１４】
【数１５】

（数１５）式を（数１４）式に代入すれば（数１６）式になり、整理すれば（数１７）式（＝数９）になる。
【００１５】
【数１６】

【００１６】
【数１７】

安定なローパスフィルタＧ_ｌｐ(ｓ)を（数１７）式の両辺に乗じて、整理すれば（数１８）式になる。
【００１７】
【数１８】

実際に計測可能な電流Ｉや端子電圧Ｖに、ローパスフィルタやバンドパスフィルタを処理した値を、下記（数１９）式のように定義する。なお、ローパスフィルタの具体例は後述する。
【００１８】
【数１９】

上記変数を用いて（数１８）式を書き直せば（数２０）式になる。
【００１９】
【数２０】

更に変形すれば（数２１）式になる。
【００２０】
【数２１】

（数２１）式は、計測可能な値と未知パラメータの積和式になっているので、一般的な適応デジタルフィルタの標準形（数２２）式と一致する。
【００２１】
【数２２】

ただし、ｙ＝Ｖ_２、 ω^Ｔ＝［Ｖ_３，Ｉ_３，Ｉ_２，Ｉ_１］
θ^Ｔ＝［−Ｔ_１，Ｋ・Ｔ_２，Ｋ，ｄ］
従って、電流Ｉと端子電圧Ｖにフィルタ処理した信号を、適応デジタルフィルタ演算に用いることで、未知パラメータベクトルθを推定する。本実施例では、単純な「最小二乗法による適応デジタルフィルタ」の論理的な欠点（一度推定値が収束すると、その後パラメータが変化しても再度正確な推定ができないこと）を改善した「両限トレースゲイン方式」を用いる。
（数２２）式を前提に未知パラメータベクトルθを推定するためのパラメータ推定アルゴリズムは下記（数２３）式となる。ただし、ｋ時点のパラメータ推定値をθ(ｋ)とする。
【００２２】
【数２３】

ただし、λ_１、λ_３(ｋ)、γ_Ｕ、γ_Ｌは初期設定値で、０＜λ_１＜１、０＜λ_３(ｋ)＜∞とする。Ｐ(０)は十分大きな値、θ(０)は非ゼロな十分小さな値を初期値とする。trace｛Ｐ｝は行列Ｐのトレースを意味する。
本発明では電流の値または電流の変化率に応じて上記の調整ゲインλ_３(ｋ）を補正する。なお、ある観測ノイズの環境下で推定精度を良好に保つには、調整ゲインには上限がある。そのため、調整ゲインを予め大きく設定しておくのは、上限までのマージンが小さくなるので、観測ノイズの影響を受けて推定精度が悪化するため得策でない。以上が、電池モデルから適応デジタルフィルタまでの導出である。
【００２３】
図４は、電子制御ユニット３０のマイコンが行う処理のフローチャートであり、第１の実施例を示す。同図のルーチンは一定周期Ｔ_０毎に実施される。例えば、Ｉ(ｋ）は今回の値、Ｉ(ｋ−１）は１回前の値を意味する。
（第１の実施例）
第１の実施例は、電池モデルの次数を１次にしたものである。
【００２４】
まず、ステップＳ１０では、電流Ｉ(ｋ）、端子電圧Ｖ(ｋ）を計測する。
ステップＳ２０では、二次電池の遮断リレーの判断する。電子制御ユニット３０は二次電池の遮断リレーの制御も行っており、リレー遮断時（電流Ｉ＝０）はステップＳ３０へ進む。リレー締結時はステップＳ４０へ進む。
ステップＳ３０では、端子電圧Ｖ(ｋ）を端子電圧初期値Ｖ_iniとして記憶する。
ステップＳ４０では、端子電圧差分値△Ｖ(ｋ）を算出する。ただし、△Ｖ(ｋ）＝Ｖ(ｋ）−Ｖ_ini
これは、適応デジタルフィルタ内の推定パラメータの初期値を約０としているので、推定演算開始時に推定パラメータが発散しないように、入力を全て０とするためである。リレー遮断時はステップＳ３０を通るので、Ｉ＝０かつ△Ｖ(ｋ）＝０なので、推定パラメータは初期状態のままである。
【００２５】
ステップＳ５０では、電流Ｉ(ｋ）と端子電圧差分値△Ｖ(ｋ）に、（数１９）式に基づきローパスフィルタＧ_ｌｐ(ｓ)、バンドパスフィルタｓＧ_ｌｐ(ｓ)及びｓ^２Ｇ_ｌｐ(ｓ)のフィルタ処理を施し、Ｉ_１、Ｉ_２、Ｉ_３、Ｖ_２、Ｖ_３を算出する。（数２３）式のパラメータ推定アルゴリズムの推定精度を良くするために、観測ノイズを低減するようローパスフィルタＧ_ｌｐ(ｓ)の応答性を遅く設定する。但し、電池の応答特性よりは速くする。ここで、下記（数２４）式のｐは、Ｇ_ｌｐ(ｓ)の応答性を決める定数である。
【００２６】
【数２４】

実際のフィルタ処理演算は、連続時間系で記述された伝達関数をタスティン近似等で離散時間化し、下記（数２５）式に示すような漸化式でフィルタ処理を行う。係数α_０〜α_３、β_１〜β_３はＧ_ｌｐ(ｓ)を離散時間化した際の定数である。
【００２７】
【数２５】

ステップＳ６０では、適応デジタルフィルタの入力である電流の振幅の大きさに応じて、（数２３）式のパラメータ推定アルゴリズムの調整ゲインλ_３(ｋ）を補正するか否かを判定する。しかし、電流は連続的に変化するため、振幅だけで調整ゲインを補正すると、パラメータ推定の最中で調整ゲインが一定でなくなり、推定精度に悪影響を与えてしまう。そのため、電流Ｉ_１(ｋ）の振幅が所定時間以上継続して所定値よりも小さい場合に、補正が必要と判断してステップＳ７０へ進む。前記条件が不成立の場合は、補正が不要と判断してステップＳ８０へ進む。なお、適応デジタルフィルタのもう一つの入力である電圧は、内部抵抗に応じて振幅が決まるので、電流の振幅だけで判断すれば十分である。
【００２８】
ステップＳ７０では、適応デジタルフィルタの入力が小さいので、通常の調整ゲインでは収束性が悪いため、λ_３(ｋ）を初期値から５倍して感度が鋭敏になるように補正する。補正の方法は、電流の振幅に応じてλ_３(ｋ）をゲインスケジューリングしても良い。
ステップＳ８０では、適応デジタルフィルタの入力が十分に大きいので、通常の調整ゲインであるλ_３(ｋ）の初期値を用いる。
【００２９】
ステップＳ９０では、ステップＳ５０で算出したＩ_１(ｋ）、Ｉ_２(ｋ）、Ｉ_３(ｋ）、Ｖ_２(ｋ）、Ｖ_３(ｋ）を（数２３）式に代入する。そしてパラメータ推定アルゴリズム（適応デジタルフィルタ演算）である（数２３）式を行い、パラメータ推定値θ(ｋ）を算出する。
ただし、ｙ＝Ｖ_２、 ω^Ｔ＝［Ｖ_３，Ｉ_３，Ｉ_２，Ｉ_１］、
θ^Ｔ＝［−Ｔ_１，Ｋ・Ｔ_２，Ｋ，ｄ］
ステップＳ１００では、ステップＳ９０で算出したパラメータ推定値θ(ｋ）の中からＴ_１、Ｋ・Ｔ_２、Ｋを用い、前記（数１４）式と等価な下記（数２６）式に基づきＧ_ｌｐ(ｓ)・Ｖ_０を算出し、これを開路電圧Ｖ_０の代用とする。開路電圧Ｖ_０は変化が緩やかなので、Ｇ_ｌｐ(ｓ)・Ｖ_０で代用できる。ただし、ここで求まるのは推定演算開始時からの開路電圧推定値の変化分△Ｖ_０(ｋ）である。
【００３０】
【数２６】

ステップＳ１１０では、ステップＳ１００で算出した△Ｖ_０(ｋ）はパラメータ推定アルゴリズム開始時からの開路電圧の変化分であるから、開路電圧初期値すなわち端子電圧初期値Ｖ_iniを加算して開路電圧推定値Ｖ_０(ｋ）を（数２７）式で算出する。
【００３１】
【数２７】

ステップＳ１２０では、図５に示す開路電圧と充電率の相関マップを用いて、ステップＳ１１０で算出したＶ_０(ｋ）から充電率ＳＯＣ(ｋ）を算出する。
なお、図５のＶ_ＬはＳＯＣ＝０％に、Ｖ_ＨはＳＯＣ＝１００％に相当する開路電圧である。
ステップＳ１３０では、次回演算に必要な数値を保存して、今回演算を終了する。
【００３２】
（第２の実施例）
第２の実施例は、電池モデルの次数を１次にしたものである。
ステップＳ１０〜ステップＳ５０は図４と同じである。
ステップＳ６０では、適応デジタルフィルタの入力である電流の変化率に応じて、（数２３）式のパラメータ推定アルゴリズムの調整ゲインλ_３(ｋ）を補正するか否かを判定する。この際、第２の実施例では、電流変化率に相当するＩ_２(ｋ）の振幅が所定時間以上継続して所定値よりも小さい場合に、補正が必要と判断してステップＳ７０へ進む。前記条件が不成立の場合、補正が不要と判断してステップＳ８０へ進む。なお、電流変化率が小さいというのは電流パターンがランプ状（塊状）の場合に相当する。
【００３３】
ステップＳ７０では、適応デジタルフィルタの入力がランプ状なので、通常の調整ゲインでは収束性が悪いため、λ_３(ｋ）を初期値から５倍して感度が鋭敏になるように補正する。補正の方法は、電流変化率Ｉ_２(ｋ）の振幅に応じてλ_３(ｋ）をゲインスケジューリング（公知技術）しても良い。
ステップＳ８０では、適応デジタルフィルタの入力がステップ状なので、通常の調整ゲインであるλ_３(ｋ）の初期値を用いる。
以下、ステップＳ９０以降は、図４と同様である。
【００３４】
次に、本出願人が先に出願した特願２００１−３８４６０６号（未公開）との関係について説明する。本発明は上記先行出願をさらに改良したものである。 上記先行出願においては、通電中の二次電池（鉛蓄電池やリチウムイオン電池等の充放電可能な電池）の端子電圧と電流に、適応デジタルフィルタを用いて開路電圧（通電遮断時の端子電圧、起電力）を推定（パラメータ同定）し、予め計測された電池の充電率と開路電圧の関係に、前記開路電圧推定値を照合して、充電率を推定する方法である。
具体的には、端子電圧Ｖと電流Ｉの関係を、開路電圧Ｖ_０および過渡項Ｔ_１、およびＫ・Ｔ２、内部抵抗Ｋを用いて、（数２８）式（＝数７）の電池モデルで表す。
【００３５】
【数２８】

（数２８）式に示す電池モデルにおいて、開路電圧Ｖ_０は初期状態からの電流Ｉの積分値に比例すると考えて、（数２９）式（＝数２）のように仮定する。
【００３６】
【数２９】

（数２９）式を（数２８）式に代入して変形すれば（数３０）式（＝数９）になり、オフセット項である開路電圧Ｖ_０を含まない。
【００３７】
【数３０】

（数３０）式を変形し、両辺に安定なローパスフィルタＧ_ｌｐ(ｓ)を乗じれば、下記（数３１）式になる。
【００３８】
【数３１】

二次電池の電流Ｉと端子電圧Ｖの計測データで、（数３１）式に公知の適応デジタルフィルタ演算（逐次型のモデルパラメータ同定アルゴリズム）を行い、式中のパラメータ（変数ｄ、過渡項Ｔ_１およびＫ・Ｔ、内部抵抗Ｋ）を一括推定することを特徴とする。
推定したパラメータ（過渡項Ｔ_１およびＫ・Ｔ_２、内部抵抗Ｋ）と電流Ｉと端子電圧Ｖを、（数２８）式と等価な下記（数３２）式に代入し、Ｇ_ｌｐ(ｓ)・Ｖ_０を開路電圧Ｖ_０として算出する。なお、ｓはラプラス演算子、Ｇ_ｌｐ(ｓ)は二次以上の安定な口ーパスフィルタである。
【００３９】
【数３２】

適応デジタルフィルタでのパラメータ推定アルゴリズムにおいて、推定精度が観測ノイズの影響を受けないように、観測ノイズを除去するためにはＧ_ｌｐ(ｓ)の応答性をかなり遅く設定する必要がある。ただし、電池の応答特性（時定数Ｔ_１の概略値は既知）よりも速い特性でないと、電池モデルの各パラメータを精度良く推定できないので、電池の応答特性よりは速く設定する。なお、上記の数式中で過渡項Ｋ・Ｔ_２を２つの変数の積で表記しているが、Ｋ・Ｔ_２で１つの変数である。
【００４０】
上記の先行出願では、リアルタイムな電流と電圧の計測値から観測ノイズの程度を判断するのは困難なので、ノイズの最大値を予め考慮してローパスフィルタＧ_ｌｐ（ｓ）の応答性を一定に設定する必要があり、それに合わせて適応デジタルフィルタでのパラメータ推定アルゴリズムに用いる調整ゲインを一定に設定している。適応デジタルフィルタヘの入力の振幅が小さい場合は、観測ノイズに備えてフィルタ処理しているため、信号成分が更に小さくなってしまい、推定精度が悪化する。しかし、上記先行出願では、調整ゲインを一定に設定するという構成になっていたため、適応デジタルフィルタの入力である電流と電圧の振幅が継続して小さい場合は、調整ゲインが不足して推定パラメータの精度が悪化する可能性がある。なお、ある観測ノイズの環境下で推定精度を良好に保つには調整ゲインには上限がある。そのため、調整ゲインを予め大きく設定しておくのは、上限までのマージンが小さくなるので、観測ノイズの影響を受けて推定精度が悪化するため得策でない。そのため本発明においては、電流（または電流の変化率）の振幅が所定時間以上継続して所定値より小さい場合に、適応デジタルフィルタ演算手段におけるパラメータ推定アルゴリズムの調整ゲインを感度大に補正するように構成している。
【００４１】
本発明においては、上記の先行出願と同様に、二次電池の電流Ｉと端子電圧Ｖの関係を示す(数１）式を、（数４）式のように近似することで開路電圧Ｖ_０（オフセット項）を含まない。そのため、計測可能な電流Ｉと端子電圧Ｖをフィルタ処理〔ｓ^ｎＧ_ｌｐ(ｓ)〕した値と、未知パラメータ〔多項式Ａ(ｓ)やＢ(ｓ)の係数パラメータおよびｄ〕の積和式が得られるので、通常の適応デジタルフィルタ（最小二乗法などで、公知のパラメータ推定アルゴリズム）を連続時間系のまま適用することが可能になる。その結果、未知パラメータを一括推定することができ、推定した未知パラメータを（数５）式に代入することで、開路電圧Ｖ_０の推定値の代用としてＧ_ｌｐ(ｓ)・Ｖ_０を容易に算出できる。これら未知パラメータは、ＳＯＣや温度や劣化度などに影響され時々刻々と変化することが分かっているけれども、適応デジタルフィルタにより精度良く逐次推定できる。そして、開路電圧Ｖ_０と充電率ＳＯＣの関係（図５）は、温度や劣化度に影響されにくく一定の関係があるため、これを予め記憶しておけば開路電圧Ｖ_０の推定値から充電率ＳＯＣを推定できる。
【００４２】
さらに、本発明においては、次のごとき特別の効果が得られる。
まず、電流の振幅が小さい場合には、パラメータ推定アルゴリズムの調整ゲインを感度鋭敏に補正するという構成であるために、電流の振幅が小さい場合でも適応デジタルフィルタで逐次推定されるパラメータの精度を良好にできる。
【００４３】
また、一般的に適応デジタルフィルタで、未知な動的システムの入出力波形からパラメータ同定を行う場合、出力波形に特徴が出易いステップ入力の方がランプ入力よりも推定精度が良い。電池の場合、入力である電流波形はモータなど負荷の動作状況に応じて多様であり、ステップとランプ（塊、集中）が混在している。そこで、電流の変化率が所定時間以上継続して所定範囲内であれば、電流がランプ状のパターンであり、電流の変化率が所定範囲外であれば、電流がステップ状のパターンであると考えることができる。ステップ状の場合には、パラメータ推定アルゴリズムの調整ゲインが通常値でもパラメータの推定精度は良好であるが、ランプ状の場合には、調整ゲインが不足するため推定精度は悪化する。しかし、請求項２においては、電流変化率の振幅が所定時間以上継続して所定値より小さい場合に、パラメータ推定アルゴリズムの調整ゲインを感度鋭敏に補正するという構成であるために、ランプ状の場合でも調整ゲインが不足しないので、パラメータの推定精度を良好にできるという効果がある。
【００４４】
以下、実際の測定例によって本発明の効果を説明する。
図６は、第１の実施例において電流が小さい場合の特性、図７は前記先行出願において電流が通常値の場合の特性、図８は前記先行出願において電流が小さい場合の特性、図９は第２の実施例において電流がランプ状態の場合の特性、図１０は前記先行出願において電流がランプ状態の場合の特性を示す図である。
【００４５】
図６〜図８は、（数７）式の電池モデルのパラメータを、温度０℃相当の値から温度２５℃相当の値に、時間４００ｓを境にステップ的に切り替えた場合に、適応デジタルフィルタでパラメータ推定したシミュレーション結果を示す図である。図８は前記先行出願で電流の振幅が通常の大きさの場合で、過渡項Ｔ_１、Ｋ・Ｔ_２、及び内部抵抗Ｋの推定値は、温度０℃から２５℃へのパラメータ変化に対応して真値に収束しているので（図中▲１▼▲２▼▲３▼）、良好な推定精度と言える。なお、過渡項Ｔ_１、Ｋ・Ｔ_２の推定値と真値に僅かに誤差があるのは、観測ノイズを電流と電圧に加算している影響である。
【００４６】
これに対し、図９は前記先行出願で電流の振幅が通常より過小の場合で、内部抵抗Ｋは良好だが（図中▲３▼）、過渡項Ｔ_１、Ｋ・Ｔ_２の推定値は、温度０℃から２５℃へのパラメータ変化に対応が遅く、真値に収束できていないので（図中▲１▼▲２▼）、推定精度が悪化している。これは、適応デジタルフィルタの調整ゲインが一定である構成のため、電流の振幅が過小な場合は、調整ゲインが不足するからである。
【００４７】
これに対して、図６に示した第１の実施例では、電流の振幅が通常より過小の場合でも、過渡項Ｔ_１、Ｋ・Ｔ_２、及び内部抵抗Ｋの推定値は、温度０℃から２５℃へのパラメータ変化に対応して真値に収束しているので（図中▲１▼▲２▼▲３▼）、良好な推定精度と言える。これは、電流の振幅が過小な場合に適応デジタルフィルタの調整ゲインを感度鋭敏に補正するという構成のため、調整ゲインに不足がないので、適応デジタルフィルタで逐次推定されるパラメータの精度を良好にできるという効果である。
【００４８】
また、図９、図１０は、（数７）式の電池モデルのパラメータを温度０℃相当の値から温度２５℃相当の値に、時間２００ｓを境にステップ的に切り替えた場合に、適応デジタルフィルタでパラメータ推定したシミュレーション結果である。図１０は前記先行出願で電流パターンがランプ状の場合で、内部抵抗Ｋは良好だが（図中▲３▼）、過渡項Ｔ_１、Ｋ・Ｔ_２の推定値は、温度０℃から２５℃へのパラメータ変化に対応が遅く真値に収束できていないので（図中▲１▼▲２▼）、推定精度が悪化している。これは、適応デジタルフィルタの調整ゲインが一定である構成のため、電流パターンがランプ状の場合は、調整ゲインが不足するからである。
【００４９】
これに対して、図９に示した第２の実施例では、電流パターンがランプ状の場合でも、過渡項Ｔ_１、Ｋ・Ｔ_２、及び内部抵抗Ｋの推定値は、温度０℃から２５℃へのパラメータ変化に対応して真値に収束しているので（図中▲１▼▲２▼▲３▼）、良好な推定精度と言える。これは、電流変化率が所定値より小さい場合（電流パターンがランプ状に相当する）に、適応デジタルフィルタの調整ゲインを感度鋭敏に補正するという構成のため、調整ゲインに不足がないので、適応デジタルフィルタで逐次推定されるパラメータの精度を良好にできるという効果である。なお、過渡項Ｔ_１、Ｋ・Ｔ_２の推定値と真値に僅かに誤差があるのは、観測ノイズを電流と電圧に加算している影響である。
【図面の簡単な説明】
【図１】本発明の一実施例を機能ブロックで表した図。
【図２】実施例の具体的な構成を示すブロック図。
【図３】二次電池の等価回路モデルを示す図。
【図４】実施例における処理のフローチャート。
【図５】開路電圧と充電率の相関マップ。
【図６】第１の実施例において、電流が小さい場合に電流Ｉと端子電圧Ｖを適応デジタルフィルタに入力して各パラメータを推定した結果を示す図。
【図７】先行出願において、電流が通常値の場合に電流Ｉと端子電圧Ｖを適応デジタルフィルタに入力して各パラメータを推定した結果を示す図。
【図８】先行出願において、電流が小さい場合に電流Ｉと端子電圧Ｖを適応デジタルフィルタに入力して各パラメータを推定した結果を示す図。
【図９】第２の実施例において、電流がランプ状態の場合に電流Ｉと端子電圧Ｖを適応デジタルフィルタに入力して各パラメータを推定した結果を示す図。
【図１０】先行出願において、電流がランプ状態の場合に電流Ｉと端子電圧Ｖを適応デジタルフィルタに入力して各パラメータを推定した結果を示す図。
【符号の説明】
１…端子電圧Ｖ（ｋ）検出手段 ２…電流Ｉ（ｋ）検出手段
３ａ、３ｂ、３ｃ、４ａ、４ｂ、４ｃ…フィルタ演算手段
７…適応デジタルフィルタ演算手段〔パラメータθ(ｋ）を推定〕
８…開路電圧演算手段〔Ｖ_０(ｋ）を演算〕 ９…充電率推定手段
１０…二次電池 ２０…負荷
３０…電子制御ユニット ４０…電流計
５０…電圧計 ６０…温度計[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an apparatus for estimating a charging rate (SOC) of a secondary battery.
[0002]
[Prior art]
The charge rate SOC (also called charge state) of the secondary battery is the open circuit voltage V₀(It is the battery terminal voltage when the power is cut off, and is also referred to as electromotive force or open voltage).₀Can be estimated. However, since the terminal voltage of the secondary battery takes time to stabilize even after the energization is cut off (end of charge / discharge), the accurate open circuit voltage V₀Is required for a predetermined time after the end of charging and discharging. Therefore, an accurate open circuit voltage V during charging / discharging or immediately after charging / discharging.₀Therefore, the charging rate SOC cannot be obtained by the above method. For this reason, conventionally, the open circuit voltage V₀Is estimated.
As a well-known example of the technology for estimating the charging rate (SOC) of a secondary battery, “Paper“ Estimation of open circuit voltage and remaining capacity of lead battery using adaptive digital filter ”, Shikoku Research Institute, Shikoku Electric Power, Yuasa Battery T.IEEE Japan Vol.112-C, No.4 1992 ”.
[0003]
[Problems to be solved by the invention]
However, in the conventional example as described above, “adaptive digital filter” is different from “non-regressive battery model (model whose output value is determined only by the current value and the past value of the input value)” which is completely different from the actual physical characteristics of the battery. The open circuit voltage is calculated using “(Sequential model parameter identification algorithm)”, and the charging rate SOC is calculated from this value. Therefore, when applied to actual battery characteristics (input: current, output: voltage), the estimation calculation does not converge at all or does not converge to a true value depending on the battery characteristics, so that an accurate charging rate SOC can be estimated. There was a problem that it was difficult.
[0004]
The present invention has been made to solve the above-described problems of the prior art, and an object of the present invention is to provide a secondary battery charge rate estimation device capable of accurately estimating the charge rate SOC and other parameters. And
[0005]
[Means for Solving the Problems]
In order to achieve the above object, the present invention is configured as described in the claims. That is, in claim 1, the battery model of the secondary battery is defined as shown in the equation (1), and the open circuit voltage V₀(Equation 2) is approximated by (Equation 2) so that (Equation 1) becomes the following (Equation 3), and adaptive digital filter operation is performed on (Equation 4) equivalent to (Equation 3). adaptive digital filter calculation means for collectively estimating the coefficient parameters of s) and B (s), and substituting these estimated values into an expression (equation 5) equivalent to the expression (equation 5)_lp(s) ・ V₀And calculate this as the open circuit voltage V₀Circuit voltage calculation means for substituting for, and previously determined circuit voltage V₀And charging rate estimation means for estimating the charging rate based on the relationship between the charging rate SOC and the adaptive digital filter calculation means if the amplitude of the current continues for a predetermined time or less and is smaller than a predetermined value. Is configured to correct the adjustment gain of the parameter estimation algorithm in FIG. The above adjustment gain is, for example, λ in the equation (Equation 23) described later.₃Corresponds to (k).
In order to improve the estimation accuracy of the parameter estimation algorithm when formulating (Equation 1) of the battery model by (Equation 4) and applying it to a known adaptive digital filter, noise is reduced in order to reduce the observation noise. The low-pass filter G considering the maximum value of_lpThe response of (s) is set to a slow constant value. Accordingly, the adjustment gain in the adaptive digital filter is also set to be constant. For this reason, when the amplitude of the current that is input to the adaptive digital filter is small, the parameter estimation algorithm may lack adjustment gain, and parameter estimation accuracy may deteriorate. However, in the present invention, when the current amplitude is small, the adjustment gain of the parameter estimation algorithm is corrected with high sensitivity. Therefore, even when the current amplitude is small, the parameter sequentially estimated by the adaptive digital filter is used. Accuracy can be improved.
[0006]
【The invention's effect】
In the present invention, the open circuit voltage V is obtained by approximating the equation (1) showing the relationship between the current I and the terminal voltage V of the secondary battery as the equation (4).₀(Offset term) is not included. Therefore, it becomes possible to apply a normal adaptive digital filter as it is in a continuous time system, so that unknown parameters can be collectively estimated, and the open circuit voltage V can be calculated by substituting the estimated unknown parameters into Equation (5).₀As a substitute for the estimate of_lp(s) ・ V₀Can be easily calculated. And open circuit voltage V₀The relationship between the charging rate SOC and the charging rate SOC (FIG. 5) is not affected by the temperature and the degree of deterioration and has a fixed relationship.₀It is possible to accurately and easily estimate the charging rate SOC from the estimated value.
Further, in the present invention, when the current amplitude is small, the adjustment gain of the parameter estimation algorithm is corrected with high sensitivity. Therefore, even when the current amplitude is small, the parameter sequentially estimated by the adaptive digital filter is used. There is a special effect that the accuracy of the can be improved.
[0007]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 is a functional block diagram showing an embodiment of the present invention. In FIG. 1, 1 is a terminal voltage V (k) detecting means for detecting a battery terminal voltage, 2 is a current I (k) detecting means for detecting a battery current, 3a, 3b, 3c, 4a, 4b, 4c are Each is a filter calculation means. 5 is adaptive digital filter calculation means [estimating parameter θ (k)], 6 is open circuit voltage calculation means [V₀(Calculate (k)), 7 is a charge rate estimating means 7 for calculating the charge rate from the open circuit voltage. The adaptive digital filter calculation means described in the claims corresponds to the combined portion of the filter calculation means 3 and 4 and the adaptive digital filter calculation means 5.
[0008]
FIG. 2 is a block diagram illustrating a specific configuration of the embodiment. This embodiment shows an example in which a secondary battery charge rate estimation device is provided in a system that drives a load such as a motor with a secondary battery or charges a secondary battery with regenerative power of the motor.
In FIG. 2, 10 is a secondary battery (also simply called a battery), 20 is a load of a motor, etc., 30 is an electronic control unit that estimates the state of charge of the battery, a CPU that calculates a program, a ROM that stores a program, and a calculation It consists of a microcomputer composed of a RAM for storing the results and an electronic circuit. Reference numeral 40 denotes an ammeter that detects current charged / discharged from the battery, 50 denotes a voltmeter that detects the terminal voltage of the battery, and 60 denotes a thermometer that detects the temperature of the battery, each connected to the electronic control unit 30. The electronic control unit 30 corresponds to the filter calculation means 3a, 3b, 3c, 4a, 4b, 4c, adaptive digital filter calculation means 5, open circuit voltage calculation means 6, and charge rate estimation means 7 in FIG. The ammeter 40 corresponds to the current I (k) detecting means 2 and the voltmeter 50 corresponds to the terminal voltage V (k) detecting means 1.
[0009]
The present invention uses an adaptive digital filter to measure the open circuit voltage V and the measured data of the terminal voltage V and current I of the secondary battery being energized.₀And the known open circuit voltage V₀2 is a device that estimates the charging rate from the relationship between the charging rate SOC and the charging rate SOC. The battery model of the secondary battery is defined as shown in the formula (1) of claim 1 and the open circuit voltage V₀Is approximated by the formula (2), the formula (1) is changed to the formula (3), and an adaptive digital filter operation is performed on the formula (4) equivalent to the formula (3). The coefficient parameters of A (s) and B (s) are collectively estimated, and these estimated values are substituted into the equation (equation 5) equivalent to the equation (equation 1) to obtain G_lp(s) ・ V₀And calculate this as the open circuit voltage V₀As an alternative to the previously described open circuit voltage V₀The charging rate is estimated based on the relationship between the charging rate and the charging rate SOC.
[0010]
The above contents will be specifically described as follows.
First, the “battery model” used in the present embodiment will be described. FIG. 3 is a diagram showing an equivalent circuit model of the secondary battery, and the battery model of the secondary battery is represented by the following equation (Equation 13).
[0011]
[Formula 13]

In the equation (13), the model input is current I [A] (positive value is charging, negative value is discharging), and model output is terminal voltage V [V], R₁[Ω] is the charge transfer resistance, R₂[Ω] is pure resistance, C₁[F] is the electric double layer capacity, V₀[V] is an open circuit voltage. Note that s is a Laplace operator. Although this model is a reduction model (primary) in which the positive electrode and the negative electrode are not particularly separated, it is possible to show the actual charge / discharge characteristics of the battery relatively accurately. As described above, in this embodiment, a configuration in which the order of the battery model is first will be described as an example.
[0012]
Derivation from the battery model of the above equation (13) to the adaptive digital filter will be described first.
When formula (13) is modified, formula (14) is obtained (= formula 7).
[0013]
[Expression 14]

T₁= C₁・ R₁, T₂= C₁・ R₁・ R₂/ (R₁+ R₂)
K = R₁+ R₂
Open circuit voltage V₀Assuming that the current I multiplied by the variable efficiency d is integrated from a certain initial state, it can be written by the equation (Equation 15) (= Equation 2 = Equation 8).
[0014]
[Expression 15]

Substituting equation (15) into equation (14) yields equation (16), and rearranging results in equation (17) (= equation 9).
[0015]
[Expression 16]

[0016]
[Expression 17]

Stable low-pass filter G_lpMultiplying both sides of (Equation 17) by (s) and rearranging results in (Equation 18).
[0017]
[Formula 18]

A value obtained by processing a low-pass filter or a band-pass filter on the current I or the terminal voltage V that can be actually measured is defined as the following equation (Equation 19). A specific example of the low-pass filter will be described later.
[0018]
[Equation 19]

If equation (18) is rewritten using the above variables, equation (20) is obtained.
[0019]
[Expression 20]

If further deformed, the equation (21) is obtained.
[0020]
[Expression 21]

Since (Equation 21) is a product-sum equation of a measurable value and an unknown parameter, it agrees with a standard form (Equation 22) of a general adaptive digital filter.
[0021]
[Expression 22]

However, y = V₂, Ω^T= [V₃, I₃, I₂, I₁]
θ^T= [-T₁, KT₂, K, d]
Therefore, the unknown parameter vector θ is estimated by using the signal filtered to the current I and the terminal voltage V for the adaptive digital filter calculation. In this example, the logical disadvantage of the simple “adaptive digital filter based on the least square method” (because once the estimated value converges, accurate estimation cannot be performed again even if the parameter changes) is improved. "Trace gain method" is used.
A parameter estimation algorithm for estimating the unknown parameter vector θ based on the equation (22) is expressed by the following equation (23). However, the parameter estimated value at the time point k is θ (k).
[0022]
[Expression 23]

Where λ₁, Λ₃(k), γ_U, Γ_LIs the default value, 0 <λ₁<1, 0 <λ₃(k) <∞. P (0) is a sufficiently large value, and θ (0) is a non-zero sufficiently small value as an initial value. trace {P} means the trace of the matrix P.
In the present invention, the above adjustment gain λ according to the value of current or the rate of change of current.₃Correct (k). Note that there is an upper limit to the adjustment gain in order to maintain good estimation accuracy under certain observation noise environments. Therefore, it is not a good idea to set the adjustment gain large in advance because the margin up to the upper limit becomes small and the estimation accuracy deteriorates due to the influence of observation noise. The above is the derivation from the battery model to the adaptive digital filter.
[0023]
FIG. 4 is a flowchart of processing performed by the microcomputer of the electronic control unit 30, and shows the first embodiment. The routine in FIG.₀Performed every time. For example, I (k) means the current value, and I (k-1) means the previous value.
(First embodiment)
In the first embodiment, the order of the battery model is first order.
[0024]
First, in step S10, the current I (k) and the terminal voltage V (k) are measured.
In step S20, the secondary battery cutoff relay is determined. The electronic control unit 30 also controls the secondary battery cutoff relay, and when the relay is cut off (current I = 0), the process proceeds to step S30. When the relay is engaged, the process proceeds to step S40.
In step S30, the terminal voltage V (k) is stored as the terminal voltage initial value V_ini.
In step S40, a terminal voltage difference value ΔV (k) is calculated. However, ΔV (k) = V (k) −V_ini
This is because the initial value of the estimation parameter in the adaptive digital filter is set to about 0, so that all the inputs are set to 0 so that the estimation parameter does not diverge when the estimation calculation starts. Since the process goes through step S30 when the relay is cut off, since I = 0 and ΔV (k) = 0, the estimation parameters remain in the initial state.
[0025]
In step S50, the current I (k) and the terminal voltage difference value ΔV (k) are converted into a low-pass filter G based on the equation (19)._lp(s), band pass filter sG_lp(s) and s²G_lpApply the filter processing of (s), I₁, I₂, I₃, V₂, V₃Is calculated. In order to improve the estimation accuracy of the parameter estimation algorithm of the equation (23), the low-pass filter G is used to reduce the observation noise._lpSet the response of (s) slower. However, it is faster than the response characteristics of the battery. Here, p in the following (Equation 24) is G_lpIt is a constant that determines the response of (s).
[0026]
[Expression 24]

In actual filter processing calculation, a transfer function described in a continuous time system is converted into discrete time by Tustin approximation or the like, and filter processing is performed using a recurrence formula as shown in the following (Equation 25). Coefficient α₀~ Α₃, Β₁~ Β₃Is G_lpIt is a constant when (s) is converted to discrete time.
[0027]
[Expression 25]

In step S60, the adjustment gain λ of the parameter estimation algorithm of equation (23) is determined according to the amplitude of the current that is the input of the adaptive digital filter.₃It is determined whether or not (k) is to be corrected. However, since the current continuously changes, if the adjustment gain is corrected only by the amplitude, the adjustment gain is not constant during the parameter estimation, which adversely affects the estimation accuracy. Therefore, the current I₁When the amplitude of (k) continues for a predetermined time or longer and is smaller than the predetermined value, it is determined that correction is necessary, and the process proceeds to step S70. If the condition is not satisfied, it is determined that correction is unnecessary, and the process proceeds to step S80. Note that the voltage, which is another input of the adaptive digital filter, has an amplitude determined according to the internal resistance.
[0028]
In step S70, since the input of the adaptive digital filter is small, the convergence is poor with the normal adjustment gain.₃(k) is multiplied by 5 from the initial value to correct the sensitivity. The correction method is λ depending on the current amplitude.₃(k) may be gain-scheduled.
In step S80, since the input of the adaptive digital filter is sufficiently large, the normal adjustment gain λ₃The initial value of (k) is used.
[0029]
In step S90, I calculated in step S50.₁(k), I₂(k), I₃(k), V₂(k), V₃Substitute (k) into Equation (23). Then, the equation (Equation 23) which is a parameter estimation algorithm (adaptive digital filter calculation) is performed to calculate the parameter estimated value θ (k).
However, y = V₂, Ω^T= [V₃, I₃, I₂, I₁],
θ^T= [-T₁, KT₂, K, d]
In step S100, T is selected from the parameter estimated values θ (k) calculated in step S90.₁, KT₂, K, and G based on the following equation (equation 26) equivalent to the equation (equation 14)._lp(s) ・ V₀And calculate this as the open circuit voltage V₀Substitute for Open circuit voltage V₀Since the change is slow, G_lp(s) ・ V₀Can be substituted. However, what is found here is the change ΔV in the estimated open circuit voltage from the start of the estimation calculation.₀(k).
[0030]
[Equation 26]

In step S110, ΔV calculated in step S100.₀Since (k) is the change in the open circuit voltage from the start of the parameter estimation algorithm, the open circuit voltage initial value V_ini is added to the open circuit voltage initial value V_ini.₀(k) is calculated by the equation (27).
[0031]
[Expression 27]

In step S120, using the correlation map between the open circuit voltage and the charging rate shown in FIG.₀The charging rate SOC (k) is calculated from (k).
Note that V in FIG._LIs SOC = 0%, V_HIs an open circuit voltage corresponding to SOC = 100%.
In step S130, numerical values necessary for the next calculation are stored, and the current calculation ends.
[0032]
(Second embodiment)
In the second embodiment, the order of the battery model is first order.
Steps S10 to S50 are the same as those in FIG.
In step S60, the adjustment gain λ of the parameter estimation algorithm of equation (23) is selected according to the rate of change of the current that is the input of the adaptive digital filter.₃It is determined whether or not (k) is to be corrected. At this time, in the second embodiment, I corresponding to the current change rate is obtained.₂When the amplitude of (k) continues for a predetermined time or longer and is smaller than the predetermined value, it is determined that correction is necessary, and the process proceeds to step S70. If the condition is not satisfied, it is determined that correction is unnecessary, and the process proceeds to step S80. A small current change rate corresponds to a case where the current pattern is in a ramp shape (lumb shape).
[0033]
In step S70, since the input of the adaptive digital filter has a ramp shape, the convergence is poor with a normal adjustment gain.₃(k) is multiplied by 5 from the initial value to correct the sensitivity. The correction method is the current change rate I₂λ depending on the amplitude of (k)₃(k) may be subjected to gain scheduling (a known technique).
In step S80, since the input of the adaptive digital filter is stepped, the normal adjustment gain λ₃The initial value of (k) is used.
Hereinafter, step S90 and subsequent steps are the same as those in FIG.
[0034]
Next, the relationship with Japanese Patent Application No. 2001-384606 (unpublished) previously filed by the present applicant will be described. The present invention is a further improvement of the above prior application. In the above prior application, the open circuit voltage (terminal voltage when the energization is cut off) is applied to the terminal voltage and current of a secondary battery that is energized (battery that can be charged and discharged such as a lead storage battery or a lithium ion battery) using an adaptive digital filter. Electromotive force) is estimated (parameter identification), and the charge rate is estimated by checking the open circuit voltage estimated value against the relationship between the battery charge rate and the open circuit voltage measured in advance.
Specifically, the relationship between the terminal voltage V and the current I is expressed as the open circuit voltage V₀And the transient term T₁, And K · T2, and the internal resistance K, the battery model of (Expression 28) (Expression 7) is used.
[0035]
[Expression 28]

In the battery model shown in Equation 28, the open circuit voltage V₀Is assumed to be proportional to the integral value of the current I from the initial state, and is assumed to be expressed by Equation (29) (= Equation 2).
[0036]
[Expression 29]

If equation (29) is substituted into equation (28) and modified, equation (30) (= equation 9) is obtained, and the open circuit voltage V, which is an offset term, is obtained.₀Not included.
[0037]
[30]

The equation (30) is modified, and a stable low-pass filter G on both sides_lpMultiplying (s) yields the following (Equation 31).
[0038]
[31]

The measurement data of the current I and the terminal voltage V of the secondary battery is used to perform a known adaptive digital filter calculation (sequential type model parameter identification algorithm) on the equation (31), and the parameters (variable d, transient term T) in the equation₁And K · T and internal resistance K) are collectively estimated.
Estimated parameters (transient term T₁And KT₂, Internal resistance K), current I, and terminal voltage V are substituted into the following (Equation 32) equivalent to (Equation 28), and G_lp(s) ・ V₀Open circuit voltage V₀Calculate as S is Laplace operator, G_lp(s) is a second-order or higher stable mouth-pass filter.
[0039]
[Expression 32]

In order to remove the observation noise so that the estimation accuracy is not affected by the observation noise in the parameter estimation algorithm with the adaptive digital filter, G_lpIt is necessary to set the response of (s) quite late. However, battery response characteristics (time constant T₁If the characteristics of the battery model are not faster than the known characteristics, each parameter of the battery model cannot be estimated with high accuracy. Therefore, the parameter is set faster than the response characteristics of the battery. In the above formula, the transient term K · T₂Is expressed as the product of two variables.₂Is one variable.
[0040]
In the above prior application, it is difficult to determine the degree of observation noise from real-time current and voltage measurement values._lpIt is necessary to set the responsiveness of (s) to be constant, and accordingly, the adjustment gain used for the parameter estimation algorithm in the adaptive digital filter is set to be constant. When the amplitude of the input to the adaptive digital filter is small, the signal component is further reduced because the filter processing is performed in preparation for the observation noise, and the estimation accuracy is deteriorated. However, in the above prior application, since the adjustment gain is set to be constant, when the amplitude of the current and voltage that are the inputs of the adaptive digital filter is continuously small, the adjustment gain is insufficient and the estimation parameter The accuracy may deteriorate. Note that there is an upper limit to the adjustment gain in order to maintain good estimation accuracy under certain observation noise environments. Therefore, it is not a good idea to set the adjustment gain large in advance because the margin up to the upper limit becomes small and the estimation accuracy deteriorates due to the influence of observation noise. Therefore, in the present invention, when the amplitude of the current (or current change rate) continues for a predetermined time or longer and is smaller than the predetermined value, the adjustment gain of the parameter estimation algorithm in the adaptive digital filter calculation means is corrected to a high sensitivity. It is composed.
[0041]
In the present invention, the open circuit voltage V is obtained by approximating the equation (1) indicating the relationship between the current I of the secondary battery and the terminal voltage V as in the equation (4), as in the above-mentioned prior application.₀(Offset term) is not included. Therefore, the measurable current I and terminal voltage V are filtered [sⁿG_lp(s)] and an unknown parameter (coefficient parameters of polynomial A (s) and B (s) and d) are obtained, so that a conventional adaptive digital filter (such as a least square method) Parameter estimation algorithm) can be applied in a continuous time system. As a result, unknown parameters can be collectively estimated, and the estimated unknown parameter is substituted into the equation (5), so that the open circuit voltage V₀As a substitute for the estimate of_lp(s) ・ V₀Can be easily calculated. Although these unknown parameters are known to change from moment to moment depending on the SOC, temperature, deterioration degree, and the like, they can be successively estimated with high accuracy by an adaptive digital filter. And the open circuit voltage V₀The relationship between the charging rate SOC and the charging rate SOC (FIG. 5) is not affected by the temperature and the degree of deterioration and has a fixed relationship.₀The charging rate SOC can be estimated from the estimated value.
[0042]
Furthermore, in the present invention, the following special effects can be obtained.
First, because the adjustment gain of the parameter estimation algorithm is sensitively corrected when the current amplitude is small, the accuracy of the parameters that are successively estimated by the adaptive digital filter is good even when the current amplitude is small. Can be.
[0043]
In general, when parameter identification is performed from an input / output waveform of an unknown dynamic system using an adaptive digital filter, estimation accuracy is better for a step input where the output waveform is more characteristic than a ramp input. In the case of a battery, the current waveform as an input varies depending on the operating condition of a load such as a motor, and steps and lamps (lumps and concentrations) are mixed. Therefore, if the current change rate continues for a predetermined time or longer and is within a predetermined range, the current is a ramp-shaped pattern, and if the current change rate is outside the predetermined range, the current is a step-shaped pattern. Can think. In the case of a step shape, the parameter estimation accuracy is good even when the adjustment gain of the parameter estimation algorithm is a normal value. However, in the case of a ramp shape, the estimation accuracy deteriorates because the adjustment gain is insufficient. However, in the second aspect, since the configuration is such that the adjustment gain of the parameter estimation algorithm is sensitively corrected when the amplitude of the current change rate continues for a predetermined time or longer and is smaller than the predetermined value, However, since the adjustment gain is not insufficient, the parameter estimation accuracy can be improved.
[0044]
Hereinafter, the effect of the present invention will be described using actual measurement examples.
6 is a characteristic when the current is small in the first embodiment, FIG. 7 is a characteristic when the current is a normal value in the prior application, FIG. 8 is a characteristic when the current is small in the prior application, and FIG. FIG. 10 is a diagram showing characteristics when the current is in the lamp state in the second embodiment, and FIG. 10 is a diagram showing characteristics when the current is in the lamp state in the prior application.
[0045]
FIGS. 6 to 8 show the adaptive digital filter when the parameter of the battery model of the formula (7) is switched stepwise from a value corresponding to a temperature of 0 ° C. to a value corresponding to a temperature of 25 ° C. at a boundary of time 400 s. It is a figure which shows the simulation result which carried out parameter estimation by (1). FIG. 8 shows a case where the amplitude of the current in the prior application is a normal magnitude, and the transient term T₁, KT₂And the estimated value of the internal resistance K has converged to the true value corresponding to the parameter change from the temperature 0 ° C. to 25 ° C. ((1), (2), (3) in the figure), so that the estimated accuracy is good. It can be said. Transient term T₁, KT₂The slight difference between the estimated value and the true value is the effect of adding observation noise to the current and voltage.
[0046]
On the other hand, FIG. 9 shows the case where the amplitude of the current is smaller than usual in the prior application, and the internal resistance K is good ((3) in the figure), but the transient term T₁, KT₂The estimated value of is slow in responding to parameter changes from a temperature of 0 ° C. to 25 ° C., and has not converged to a true value ((1) (2) in the figure), so the estimation accuracy is deteriorated. This is because the adjustment gain of the adaptive digital filter is constant, so that the adjustment gain is insufficient when the amplitude of the current is too small.
[0047]
On the other hand, in the first embodiment shown in FIG. 6, even when the amplitude of the current is smaller than usual, the transient term T₁, KT₂And the estimated value of the internal resistance K has converged to the true value corresponding to the parameter change from the temperature 0 ° C. to 25 ° C. ((1), (2), (3) in the figure), so that the estimated accuracy is good. It can be said. This is because the adjustment gain of the adaptive digital filter is sensitively corrected when the amplitude of the current is too small, so there is no shortage of the adjustment gain, so the accuracy of the parameters that are sequentially estimated by the adaptive digital filter is improved. It is an effect that can be done.
[0048]
FIG. 9 and FIG. 10 show the adaptive digital when the parameter of the battery model of (Equation 7) is switched from a value corresponding to a temperature of 0 ° C. to a value corresponding to a temperature of 25 ° C. step by step for a time of 200 s. It is the simulation result which carried out parameter estimation with the filter. FIG. 10 shows a case where the current pattern is a ramp shape in the prior application and the internal resistance K is good ((3) in the figure), but the transient term T₁, KT₂The estimated value of is slow to respond to a parameter change from a temperature of 0 ° C. to 25 ° C. and has not converged to a true value ((1) (2) in the figure), and therefore the estimated accuracy is deteriorated. This is because the adjustment gain of the adaptive digital filter is constant, so that the adjustment gain is insufficient when the current pattern is ramp-shaped.
[0049]
On the other hand, in the second embodiment shown in FIG. 9, even if the current pattern is a ramp, the transient term T₁, KT₂And the estimated value of the internal resistance K has converged to the true value corresponding to the parameter change from the temperature 0 ° C. to 25 ° C. ((1), (2), (3) in the figure), so that the estimated accuracy is good. It can be said. This is because the adjustment gain of the adaptive digital filter is corrected sensitively when the current change rate is smaller than the predetermined value (the current pattern corresponds to a ramp), so there is no shortage of adjustment gain. The effect is that the accuracy of the parameters sequentially estimated by the digital filter can be improved. Transient term T₁, KT₂The slight difference between the estimated value and the true value is the effect of adding observation noise to the current and voltage.
[Brief description of the drawings]
FIG. 1 is a functional block diagram illustrating an embodiment of the present invention.
FIG. 2 is a block diagram showing a specific configuration of the embodiment.
FIG. 3 is a diagram showing an equivalent circuit model of a secondary battery.
FIG. 4 is a flowchart of processing in the embodiment.
FIG. 5 is a correlation map between an open circuit voltage and a charging rate.
FIG. 6 is a diagram illustrating a result of estimating each parameter by inputting a current I and a terminal voltage V to an adaptive digital filter when the current is small in the first embodiment.
FIG. 7 is a diagram showing a result of estimating each parameter by inputting a current I and a terminal voltage V to an adaptive digital filter when the current is a normal value in the prior application.
FIG. 8 is a diagram showing a result of estimating each parameter by inputting a current I and a terminal voltage V to an adaptive digital filter when the current is small in the prior application.
FIG. 9 is a diagram showing a result of estimating each parameter by inputting a current I and a terminal voltage V to an adaptive digital filter when the current is in a lamp state in the second embodiment.
FIG. 10 is a diagram showing a result of estimating each parameter by inputting a current I and a terminal voltage V to an adaptive digital filter when the current is in a lamp state in the prior application.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Terminal voltage V (k) detection means 2 ... Current I (k) detection means
3a, 3b, 3c, 4a, 4b, 4c ... filter operation means
7: Adaptive digital filter calculation means [estimates parameter θ (k)]
8: Open circuit voltage calculation means [V₀(k) is calculated] 9... Charging rate estimation means
10 ... secondary battery 20 ... load
30 ... Electronic control unit 40 ... Ammeter
50 ... Voltmeter 60 ... Thermometer

Claims (4)

The current I and the terminal voltage V of the secondary battery are measured, the open circuit voltage V ₀ is estimated from the measured values of the current I and the terminal voltage V using an adaptive digital filter, and the previously obtained open circuit voltage V ₀ and the charge are charged. A charging rate estimation device that estimates a charging rate based on a relationship with a rate SOC,
The battery model of the secondary battery is defined as shown in the following (Expression 1), and the open circuit voltage V ₀ is approximated by the following (Expression 2) to change the (Expression 1) to the following (Expression 3). Adaptive digital filter operation means for performing adaptive digital filter operation on the following equation (equation 4) equivalent to equation (3) and collectively estimating coefficient parameters of A (s) and B (s);
By substituting these estimated values into the following (Equation 5) equivalent to the (Equation 1), G _lp (s) · V ₀ is calculated, and the open circuit voltage calculation means substituting this for the open circuit voltage V ₀ ; ,
Charging rate estimating means for estimating the charging rate based on the relationship between the open circuit voltage V ₀ and the charging rate SOC determined in advance;
In addition, when the amplitude of the current continues for a predetermined time or longer and is smaller than a predetermined value, the adjustment gain of the parameter estimation algorithm in the adaptive digital filter calculation means is corrected to a high sensitivity, and the secondary battery is charged Rate estimation device.

In the above equations, A (s) and B (s) are expressed by the following equation (6).

s is a Laplace operator, A (s) and B (s) are polynomial functions of s, the order is n, d is a variable, and G _lp (s) is a transfer function having a low-pass filter characteristic. Have. The current I and the terminal voltage V of the secondary battery are measured, the open circuit voltage V ₀ is estimated from the measured values of the current I and the terminal voltage V using an adaptive digital filter, and the previously obtained open circuit voltage V ₀ and the charge are charged. A charging rate estimation device that estimates a charging rate based on a relationship with a rate SOC,
The battery model of the secondary battery is defined as shown in the following (Expression 1), and the open circuit voltage V ₀ is approximated by the following (Expression 2) to change the (Expression 1) to the following (Expression 3). Adaptive digital filter operation means for performing adaptive digital filter operation on the following equation (equation 4) equivalent to equation (3) and collectively estimating coefficient parameters of A (s) and B (s);
By substituting these estimated values into the following (Equation 5) equivalent to the (Equation 1), G _lp (s) · V ₀ is calculated, and the open circuit voltage calculation means substituting this for the open circuit voltage V ₀ ; ,
Charging rate estimating means for estimating the charging rate based on the relationship between the open circuit voltage V ₀ and the charging rate SOC determined in advance;
And when the amplitude of the rate of change of the current continues for a predetermined time or longer and is smaller than a predetermined value, the adjustment gain of the parameter estimation algorithm in the adaptive digital filter calculation means is corrected to a high sensitivity. Battery charge rate estimation device.

In the above equations, A (s) and B (s) are expressed by the following equation (6).

s is a Laplace operator, A (s) and B (s) are polynomial functions of s, the order is n, d is a variable, G _lp1 (s) is a transfer function having a low-pass filter characteristic, and an order of n + 1 or more is given. Have. The secondary battery charging rate estimation apparatus according to claim 1 or 2, wherein the order of the battery model is first order. The following (Expression 7), (Expression 8), and (Expression 9) respectively corresponding to the above (Expression 1), (Expression 2), (Expression 3), (Expression 4), and (Expression 5). The charging rate estimation of the secondary battery according to claim 1 or 2, wherein the order of the battery model is made first order by using the formula (10), the formula (10), and the formula (11). apparatus.

_{However, T 1, K · T 2} , K respectively battery _{parameters, I 1 ~I 3, V 1} ~V 3 is represented by the following equation (12) below.

Images