JP4454001B2 - Remote electrical equipment monitoring method and apparatus, and power consumption estimation method and apparatus using the same - Google Patents

Description

そのうち、８機器（表１末尾欄参照）のスイッチのオン／オフを変えた２５６（＝2⁸）ケースについて、約５分間（９００秒）の計測を行っている。他の７機種中、ポットと冷蔵庫の２機器は、自動運転により自律的に保温・冷却モードに入ったが、電子レンジなど５機器はオフ状態で接続したため、消費電流はほとんどない。なお、白熱灯は同型が５球あるが、これをオンとするケースでは、時間経過に従い１個、２個、…、５個と順次オンにしていった。蛍光灯やインバータ蛍光灯についても同様の操作をしている。
〔電気機器波形の特徴と電流変化値の独立性仮説〕
独立性の仮定に基づく信号分離を行うためには、分離結果となる各波形が互いに独立、すなわち、関連をもたずに変化する必要がある。しかし、この仮説は電気機器の電流波形については必ずしも成立しない。
【００６４】
一例として、エアコン２台（A1,A2）、インバータ式蛍光灯（F1）、テレビ１台（T3）をオンとしたケース１９７（11000101）について、図７に示す。図７の上段は、各機器の5分間（900秒間）の実効総電流値のグラフである。
【００６５】
エアコンA1、A2が類似した形で変化していることがわかる。これは、両方のエアコンをほぼ同時に起動した結果である。このため、独立性の仮説に基づき分離すると、両エアコンの共通挙動の成分と差分成分とが各独立成分となる。ほぼ同時に複数の機器のスイッチをつけるようなことはしばしば見受けられるので、電気機器の電流値間に独立性を直接仮定することは難しい。
【００６６】
一方、図７の下段は、各時刻の各機器の実効電流値の変化量（時刻ｔと時刻ｔ＋１での電流値の差、以下、電流変化（量））を示す。各機器の電流変化量の時間変化に類似性が少なく独立性が高いことがわかる。機器の動作の全体的傾向が類似している場合でも、エアコンが冷却動作に入る（実効電流値の立ち上がり部）など、各機器が特定の動作モードに入るタイミングは特別な同期機構がない限り、各機器で自律的に制御される。このため、電流変化の独立性が高くなっていると推測できる。
【００６７】
各機器の電流変化の独立性は、各機器動作の自律性を反映したものであり、一般的に成立する有効な仮説であると期待できる。なお、電流変化から求めた分離行列はそのまま実効電流に対する分離行列として使用できる。
【００６８】
独立性仮説の検証のために、稼動機器の実効電流に対して標準的な独立成分分析アルゴリズムJADE（ J.F. Cardoso et. al.: “Blind beamforcing for non-Gaussian signals”, IEE Proceedings -F, 140(6):362-370, 1993.）を適用した場合と、稼動機器の電流変化に対して同じアルゴリズムを適用した場合との比較を、全２５６ケースについて行った。その結果を図８に示す。この実験では、各稼動機器の実効電流、電流変化を分散１に正規化して入力信号として与えているので、理想的には入力信号そのものが独立成分となる。すなわち、分離行列は単位行列となるべきである。図８は、求められた分離行列と単位行列とのずれ（対角成分の絶対値の最小値と１との差）を示す。０は完全な分離・再現ができていることを、１に近くなるほど、分離がうまくいかず混信していることを意味する。
【００６９】
電流変化による場合は、混信の度合いは最大０．５７で、２５６ケース中２１７ケースで混信の度合いが０．０５以下の精度の高い分離を実現しているのに対して、実効電流による場合には、０．９以上の混信が半数以上の１３２ケースで生じている。
【００７０】
各機器の電流変化の独立性仮説は、機器の実効電流の分離に有効であることが確認された。
【００７１】
〔機器別実効電流と線形性仮説〕
次に、計測される総電流の各周波数の実効電流と推定したい各機器の実効電流の関係について検討する。
【００７２】
図９は、ケース１９７での各周波数成分およびエアコン（A1、A2）の電流変化（量）の時間変化である。エアコンA1とA2が冷却モードから定常運転に切替わった時点（図９の時刻150秒付近の点線部）に着目すると、両者で第３次波が同じように発生するのに対して、第5次波では逆向きに、また、第13次波ではエアコンA2のみが強く影響していることが観察できる。このように、各機器の動作モードの変化は、各周波数成分の電流変化値に異なる寄与をしている。従って、各周波数成分の電流変化から各機器の電流変化を推定できる可能性がある。ここで問題となるのは、独立成分分析で要請される「線形性」の仮定、すなわち、各信号源の波形が各観察信号の波形の線形合成であらわされるという性質が成立しているか否かである。
【００７３】
各機器からの電流の総和が総電流であるから、本仮説は、機器iの各周波数での実効電流（Ij）の比は、機器固有で、出力レベルや時間によらず一定、すなわち、
（I₁(t), I₃(t),…,I₁₃(t)）＝(a₁, … ,a₁₃) ・I₀(t)
となることを意味する。「機器の各周波数での実効電流の比は、時間的に変化せず、機器固有の値を持つ」という本仮説は、独立性仮説と同様に厳密には成立しない。しかし、近似的に成立すれば、総電流の各周波数成分の実効電流（電流変化）から、機器電流の電流変化の概略が推定可能となる。
【００７４】
実験では、各機器単独での基本周波数の実効電流は計測されているが、周波数別実効電流は計測されていないので、上記仮説の直接的検証は困難である。ここでは、総電流の周波数別実効電流を用いて上記仮説が成立したと仮定した場合の機器の基本周波数での実効電流の予測精度を調べることで、上記仮説の成立状況を検証する。
【００７５】
各周波数成分の実効電流値と機器の実効電流値から、誤差の２乗平均の最小化（最小二乗法）によって、分離行列Bを求めた結果を図１０に示す。図１０の上段は、各周波数での実効電流値から分離行列Bによって得た各機器の近似実効電流値である。図１０の下段は、各機器の近似実効電流値の差分値（以下、近似差分電流）である。図７と見比べると、エアコン（A1,A2）などについては、近似実効電流は実際の実効電流に近く、線形性が近似的に成立していることがわかる。一方、インバータ蛍光灯（F1）とポット（J）については、近似波形が実際の実効電流と大きく異なり、線形性の仮定は良い近似を与えていない。ただし、両者の合計値（F1＋J）については、各々単独に推定するより近似精度が改善し、近似的に線形性が成立していると考えられる。これは、近似差分電流値（図１０）の各波形類似性から推察されるようにインバータ蛍光灯とポットとは類似した周波数特性（各周波数での実効電流の比率(a1/a1,..,a13/a1)）をもつためと考えられる。
【００７６】
従って、線形性の仮定は、個別機器について必ずしも成立するものではないが、類似した周波数特性をもつ特定機器群を一つの機器とみなせば、近似的に成立する。インバータ蛍光灯とポットのような類似した周波数特性を示す機器をさらに分離するためには、各機器の基本波の強度レベルI₁＝a1・I₀に関する仮説、すなわち、機器固有の電流変化の値（例：ポットJ:0．6A）を導入する必要がある。（後述するように、これには、機器動作特性モデルを使用すれば良い。）
【００７７】
周波数特性に基づく機器グループを検討するために、各ケースの動作機器間の近似差分電流、および、近似差分電流の近似誤差（近似差分電流−実際の電流変化）の相関係数を求め、絶対値平均をとったものを図１２に示す。前節で論じたように実際の電流変化自身は、独立性が高く強い相関を示さない。近似差分電流の相関係数が大きいことは、当該機器同士が類似した周波数強度比を持つことを意味し、両機器の電流変化が一つの独立な信号成分として分離されやすいと予測される。一方、近似電流変化誤差の相関係数は、電流変化のうち線形性で捉えきれない部分の挙動の類似性を示す。従って，この相関が高い機器は、別々の独立信号成分として分離されるが、特定の時刻で両者の混信が生じる傾向があることと予測される。
【００７８】
図１２から、機器群｛白熱灯(4)、蛍光灯(5)、インバータ式蛍光灯(6)、ポット(9)｝は、近似差分電流の相関係数(線形性部分)が高く、独立成分分析では区別しづらい機器群であると予測できる。また、機器群｛エアコンA2(2),扇風機(3),テレビ１(7)｝も線形性に関する類似性が高い。エアコン（A1(1)，2(2)）は、扇風機(3)とテレビ１(7)以外のすべての機器との非線形な類似性を持つことがわかる。特に、エアコン相互の関連性が強く、部分的混信が生じやすいと考えられる。
【００７９】
〔機器別電流変化の推定〕
〔標準的アルゴリズムの適用実験〕
機器別電流変化に関する独立性、および、適当な機器グループに対する線形性を仮定し得ることが確認された。本節では、各周波数成分の毎秒の実効電流値のみから、上記の独立性と線形性の仮定に基づいて、未知の接続機器の毎秒の電流変化を推定する実験について述べる。
【００８０】
実験では、まず、標準的独立成分分析手法JADEを適用して個別機器の電流変化を推定した。機器動作推定においては、一種の例外値とも言えるオン・オフ時などのスパイク的電流変化が重要となるが、JADEは、例外値の影響が強い独立性指標クルトシスの平方和の最大化を効率的に行う。このため、JADEアルゴリズムを選択した。
【００８１】
図１２にケース１９７での基本波〜１３次（奇数次）までの７周波数（図９の１段〜７段）を入力として信号源の波形（基本周波数の電流変化）を推定した結果を示す。ｊ次高調波の電流変化X_ｊ、混合行列A、分離行列Bとする時、第i推定成分の基本波の電流変化はS_i＝Σ_jA(1,i)・B(i,j)・X_jにより求まる。第１成分(最大振幅3A),第２成分(2.8A),第３成分(0.8A)，第４成分(0.4A)が主な成分である。図１３に、第１〜第３成分と、対応する稼動機器の波形（基本波の電流変化）を示す。
【００８２】
第１成分はエアコンA１を、第２成分はエアコンA２を比較的良く再現している。ただし、詳しく見ると第１成分には、エアコンA2の起動時の成分が一部混信している（３００秒、６００秒付近）。なお、第1、２成分の周波数強度は、混合行列Aを調べると、基本波、３次波、５次波が主で第１成分が約１０：６：２に、第２成分が１０：５：−１の強度比となっている。
【００８３】
第３成分は、単独の機器に対応していない。インバータ蛍光灯の点灯（４回、約０．２A），およびポットの起動・停止（５回、約±０．７A）に対応したスパイク成分と，エアコンA2の一部（停止時の小スパイク（３回，−０．５A）と起動時の一部）が混合している。図示していないが、実験結果から、第４成分（スパイク２回、０．４A）は、エアコンA2の２回、３回目の起動時ピークに対応している。第５成分はテレビとの相関が強い。
【００８４】
第３成分では、基本波と３次波の強度比が１０：１で他の周波数の強度は無視できる。すなわち、ほとんど高調波を伴わない機器成分である。ポット、インバータ蛍光灯が第３成分に含まれることは、前述のインバータ蛍光灯F1とポットJの類似性の議論と一致する。エアコンが一時的に第３成分に混信するのは、エアコンのインバータ制御により一時的に高調波が弱くなったためと考えられる。エアコンA2については、特に起動時の波形が，対応する第２成分以外の他成分に混信し、また、一部のスパイクが第4成分として分離される。これは、エアコンのインバータ制御による非線形性が原因と考えられる。この点も前述の議論と合致する（図１１の(b)参照）。
【００８５】
上記の実験から、同一周波数強度特性や非線形性による一部混信などがあるが、総電流の奇数次周波数成分の実効電流値から、主要な各機器の動作状況を、ほぼ再現できることが示された。
〔機器固有モデルによる精度向上〕
一般の独立成分分析では、分離対象機器の特性は完全に不明であるとして、各信号源ｓ_iの性質とは無関係に同一の独立性指標H(ｓ)を仮定して、その総和Σ_i H(ｓ_i ) が最小になるように分離行列を決定する。
【００８６】
しかし、定常的な監視を行う信号分離追跡問題、特に、機器動作信号分離においては、接続機器の種類が事前にある程度想定でき、その定格などから電気機器特性を事前に把握できる場合がある。事前には不明でも、信号分離を定常的に進めるなかで同一特性を持つ信号源として特定される場合もある。このような場合、機器固有の動作特性を利用することで、分離精度を向上できる。
【００８７】
提案手法では、機器動作特性を当該既機器の電流変化（基本波）の標準的な電流変化値で与える。たとえば、オン・オフ型機器では、オン動作時の電流変化値、オフ動作時の電流変化値はほぼ一定値を取る。
【００８８】
提案手法では,これらの標準的分布Qi(x)が与えられると、信号源特性に基づく独立成分分析例えば図４のアルゴリズムを含めた独立成分分析によって、適切な独立性指標H(s)＝Σ_iH(s_i)を求めてその最大化を行う。
【００８９】
図１５にエアコン等の標準的確率分布を与えた時に得られた最適な独立性指標（主因子）と代表的手法で使用される独立性指標とを示す。JADEが良い分離性能を示すのは、他の独立性指標に比べ、JADEで使用する独立性指標（x⁴）が最適な独立性指標の良い近似となっているためと考えられる。
【００９０】
提案アルゴリズムでは,さらに、与えられた機器動作特性モデルに基づいて、分離された独立成分に生じている混信の解消を行う。
【００９１】
ケース197の例では,第3成分に、インバータ蛍光灯（F1）とポット（J）とエアコン（A2）の一部の混信が見られるが、これは、インバータ蛍光灯、ポット各動作特性モデルQi(s) （蛍光灯：定常状態の０Ａと点灯時の約０．２Ａに二つのピークを持つ混合分布、ポット：定常状態０Ａと、起動・停止時の約±０．６５Ａにピークを持つ混合分布）とによって分離される。
【００９２】
図１６に、機器固有モデルと拡張した独立性分析アルゴリズムを用いて最終的に推定された機器電流変化を示す。機器固有の動作モデルの組み込みにより、特に、高調波の少ない機器の分離精度が改善していることがわかる。
【００９３】
以上により,提案する機器動作の独立性に着目した提案手法が,機器動作分離追跡に有効であることが示された。さらに、10台の稼動家電機器で計測したデータに対して提案手法を適用することにより、個別機器に計測装置、情報収集装置を設置せず、また、接続されている機器の動作特性に関する情報がなくても、主要な機器の動作状態を推定できることを示した。
【００９４】
また、機器動作特性モデルが与えられる時には、混信して一つの機器の電流変化として認識されている電流変化も、異なる機器の電流変化として認識するなど、推定精度を向上できる点を示した。
【００９５】
機器動作信号分離問題では、機器の独立な動作という性質が、機器の電流変化（電流変化）に現れることに着目することで、主要接続機器の電流変化（電流変化）に対応する成分を分離することができた。この性質は、機器が非同期に自律的に動作するという非常に一般的な性質に立脚している。
【００９６】
ただし、線形性仮説、すなわち、各機器の高調波の強度比は、機器固有で時間的にほぼ一定であるという性質は厳密には成立しない。これには、２つの原因がある。一つには,インバータ蛍光灯とポットのように、類似の強度比を持つ機器がある点、エアコンなどのように強度比が時々変動するという非線形性を持つ点である。第１の原因により,類似強度比を持つ複数機器の同一化が生じ、第２の原因により、他の成分への混信や、単一機器の複数成分への分離といった信号の分散が生じる。これらが推定精度低下を引き起こしている。
【００９７】
類似強度比を持つ機器については,機器固有の動作特性モデルが与えられるならば、そこから得られる強度情報に基づいて、提案手法により別機器の電流変化として認識できることを示した。非線形性に伴う信号の分散については、対応する信号成分（ケース１９７のエアコン２では第3成分）が不明になるほどの激しい分散ではなく、動作状態推定には有効なレベルにとどまっている。しかし、推定精度の向上および小電流機器の動作状態推定を行うには、対処が必要となる。
【００９８】
独立性に着目して、接続機器の動作を推定する技術は、機器固有のモデルを事前に用意しなくてもある程度動作推定が行える点で非常に有用な技術である。
【００９９】
なお、上述の実施形態は本発明の好適な実施の一例ではあるがこれに限定されるものではなく本発明の要旨を逸脱しない範囲において種々変形実施可能である。例えば、本実施例においては各機器の基本波の持つ実効電流値、即ち実効電流値Ｉ₀(t)の時間差分δＩ(t)を使用する方法を述べているが、各次数の高調波次数の電流と電圧の位相差ω(t)により、各次数の高調波の電流値を
Ｉ(t)＝√2・Ｘ₀(t)・(cos(ω(t))+ｉ・sin(ω(t)))
として複素表現 (ｉ：虚数)をすることで、無効電流を含めた電流に対しても容易に拡張可能である。
【０１００】
総電流値の各次数の高調波電流値（複素表現）の時間差分δＩ(f,t)に対して、複素数行列に対する独立成分分析手法を使用することで、同様に、各機器の基本波の電流値（複素表現）の時間差分δS(i,t)を推定することができる。この手法は、位相差ω(t)をほぼ一定と近似した場合に相当する。
【０１０１】
また、本実施形態では主に非侵入的な電力機器の個別の消費電力の推定について述べたが、特に利用方法は限定されず、電気機器の動作異常を警告することにも利用できる。即ち、電気機器の消費電力推定システムで得られた電力消費に関する情報から、例えば日常の電力消費との比較において異常と判断される場合に、電力需要家在室者の安否、電力需要家内の安全、電気機器や電化システムの異常の有無等を判定し、その情報を外部へ発信することができる。例えば、本システムにより、在室者が毎日オンオフされるはずの照明、テレビ、電気ポット、温水便座等の動作状態から「電力需要家在室者の安否」を判定することができるとともに、火災等の原因となる電気アイロン、電気ストーブ、電化厨房等の長時間使用（つけっぱなし）等から「電力需要家内の安全」を判定することができる。「これらの情報の外部発信」については、既存の電話回線、PHS、ポケベル、インターネット等の利用が可能であり、「通報対象者」は居室者本人、居室者の縁者、消防署、地方自治体等の福祉医療担当者等を想定できる。
【０１０２】
【発明の効果】
以上の説明より明らかなように、請求項１及び４記載の本発明の遠隔電気機器監視方法及び装置によると、ある機器の電流変化が他の機器の電流変化とは独立しているということに着目し、独立成分分析手法を利用して総消費電流から同一の高調波強度比率成分の機器別の電流変化を分離し、機器別に可動状況を推定することができるので、事前にデータベースを構築しなくとも複数の機器の動作状況を把握することができる。データベースを構築する必要がないので、低コストでかつ簡便に複数機器の総消費電流値のみから個別機器の消費電流やその変化、動作状態を簡便に推定できる。しかも、需要家の給電線の任意の１点、例えば屋外の給電線引込口付近に測定センサーを設置するだけで、被測定電気機器毎に測定センサーを取り付ける必要がないので、本システムを電力需要家に設置するときにプライバシー等を侵害したり、追加の配線等を施す度合いが少ない利点がある。
【０１０３】
そして、このような需要家（工場、ビル、一般家庭）での電気機器の実際の使用状況の把握は、電気事業にとっても需要家にとっても重要である。電力においては、料金システムの構築、需要家への各種省エネルギーサービス事業の展開に、需要家においては省エネ運転制御や、機器故障の検出などに活用できる。
【０１０４】
また、請求項２記載及び５の発明によると、同一高調波比率を有する機器群の電流変化も、機器毎の情報例えば基本波の電流変化（定格情報）に基づき更に分離することができるので、電流変化量の推定精度を高めることができる。
【０１０５】
更に、請求項３並びに６記載の発明によると、同一高調波比率を示す機器群毎の消費電流並び消費電力を求めることができる。
【図面の簡単な説明】
【図１】本発明の遠隔電気機器監視方法の一実施形態を示す概要図である。
【図２】本発明の遠隔電気機器監視装置の一実施形態を示す概略ブロック図である。
【図３】周波数起用度比が同じ機器群別の基本波電流変化量の推定工程の一例を示すフローチャート。
【図４】最適独立性指標の推定工程の一例を示すフローチャート。
【図５】最適独立性指標Ｇを用いた時の分離行列更新アルゴリズム（図３の推定工程の第４工程）の一例を示すフローチャート。
【図６】機器群基本電流変化量からの機器別電流変化量の推定工程の一例を示すフローチャート。
【図７】各機器の実効電流と電流変化の変化（ケース１９７）のグラフである。
【図８】機器の実効電流による分離と電流変化に基づく分離の混信の度合いを示すグラフである。
【図９】各周波数での電流変化と、エアコン（A1,A2）の電流変化値を示すグラフである。
【図１０】機器別近似実効電流と近似差分電流とを示すグラフである。
【図１１】近似差分電流と近似差分電流誤差の相関係数の絶対値（全ケース平均）を示すグラフで、（ａ）は近似差分電流の相関を、（ｂ）は近似差分電流誤差の相関をそれぞれ示す(各軸の数字は機器番号）。
【図１２】本発明の独立成分分析の結果である周波数別実効電流変化に基づく信号分離結果（ケース１９７）のグラフである。
【図１３】第１，第２，第３独立成分と稼動機器の電流変化の比較図である。
【図１４】エアコンA2の動作特性モデル（電流変化の取りえる確率の対数）を示すグラフである。
【図１５】代表的独立性指標と最適化された独立性指標を示すグラフである。
【図１６】動作特性モデルを用いて更に独立成分分析の結果である推定電流変化（括弧内：対応稼動機器）を示すグラフである。
【符号の説明】
１ 電力需要家
２ 給電線
３ 電気機器
１１ 測定センサー
１２ 周波数成分変換装置
１３ 時間差分装置
１４ 独立成分分析装置
１５ 機器別信号分離装置[0001]
BACKGROUND OF THE INVENTION
The present invention relates to remote electricity that can estimate the movement status and current consumption of individual devices based on the total current consumption of the current consumption of a plurality of electrical devices used by electric power consumers (electric users). The present invention relates to an apparatus monitoring method and apparatus, and a power consumption estimation method and apparatus using the apparatus monitoring method and apparatus. More specifically, the present invention relates to a remote electric device monitoring method and apparatus suitable for estimating the movement status (current consumption) and power consumption of a plurality of electric devices by a non-intrusive method, and consumption of the electric devices. The present invention relates to a power estimation method and apparatus.
[0002]
[Technical terms]
In this specification, “non-intrusive” means that a measurement sensor is installed at one location in the vicinity of the feeder inlet, and an electrical sensor connected to the circuit is attached to each branch circuit downstream of the feeder line. This means that no measurement sensor is attached to each device. Moreover, an inverter apparatus means what can carry an inverter and can change operation | movement of an apparatus continuously from a low output to a high output. The power consumption of this inverter device changes continuously from a small value to a large value according to the output. Furthermore, the non-inverter device refers to a device that does not have an inverter and is in a state where the operation of the device is simply on and off. The power consumption of the non-inverter device takes a limited value corresponding to the on / off operation.
[0003]
[Prior art]
Monitoring the operating status of each device in a factory or home is indispensable for promoting efficient use of power devices. A reliable method for grasping the operation status of a plurality of devices is to collect information by installing an operation status measurement device and a measurement information transmission device in each device. However, the installation of measuring devices and transmission devices becomes a factor of high costs when the number of target devices increases. In some cases, the measuring device or the transmission device cannot be installed due to environmental conditions. In particular, it is difficult for electric power consumers such as ordinary households to install sensors in the house. For this reason, a monitoring technique for estimating the operation status of the device such as the current consumption of the individual device and its change from the total current consumption of the device without installing such an individual measuring device or information collecting device is desired. Yes.
[0004]
As a monitoring system for non-intrusive estimation of the operating state of such an electrical device, an EPRI (Electric Power Research Institute; USA) apparatus has been conventionally used by using an algorithm developed by MIT (Massachusetts Institute of Technology; USA). There is something that has become. This monitoring system regards the on / off operation of electrical equipment as a step-like time change of the total power load curve of the power consumer, and electrical equipment that is turned on or off based on the rated power consumption and power factor of the electrical equipment. Identification and operation state estimation. Therefore, it is possible to identify the electrical device that performs a simple on / off operation and estimate the operation state.
[0005]
[Problems to be solved by the invention]
However, recently, inverter devices such as air conditioners have become widespread in ordinary households, and non-inverter devices and inverter devices are often used in a mixed state. Since the inverter device controls the output according to the state of the load, the power consumption also changes accordingly. Therefore, the temporal transition of the power consumption is not necessarily stepped, and varies slowly or irregularly.
[0006]
Therefore, in a situation where inverter devices and non-inverter devices coexist, it is difficult to estimate the power consumption of each individual electric device depending on the conventional EPRI development monitoring system described above, and the estimation of the operating state of the electric device. Even difficult.
[0007]
In order to solve this problem, the applicants of the present application, etc. as a technique for estimating the operating status of a plurality of electrical devices including inverter devices and non-inverter devices, Based on the harmonics of the current, an estimation algorithm such as a large margin classifier was used in advance to propose a system that estimates the operating state for each device from the effective current and phase of each harmonic of the total current (Japanese Patent Application 2000). -111271).
[0008]
However, with this method, it is necessary to prepare in advance a large number of harmonic data of the total current when operating in each operating state for the assumed monitoring target device, and target completely unknown monitoring target devices. I can't do it. For this reason, if the measured total current includes an unknown device to be monitored, the unknown electric device or group of electric devices is operating and how much power is consumed. Since it can only be comprehensively estimated, if the number of unknown electrical devices increases, it becomes practically difficult to monitor the operating status of the electrical devices. Therefore, when there are various types of electrical devices that are not assumed in advance or specific monitoring target devices for each power consumer, or when replacement of target devices occurs, the operation status of many devices It is necessary to prepare the data in advance and remake the operation state determination system by learning the estimation algorithm.
[0009]
The present invention relates to a remote electric device monitoring method capable of non-intrusive estimation of individual operation states of a plurality of electric devices used by a power consumer in a situation where non-inverter devices and inverter devices are mixed, and It is an object of the present invention to provide an apparatus and a power consumption estimation method and apparatus using the apparatus. Furthermore, the present invention provides a remote electrical device that can estimate individual current consumption of a plurality of devices from only the total current without collecting data in advance and configuring an operation state determination system for each monitored device. The purpose is to provide a monitoring system.
[0010]
[Means for Solving the Problems]
In order to achieve such an object, the present inventor paid attention to the fact that the change in the consumption current of the electrical equipment installed in the power consumer fluctuates independently from the change in the consumption current of the other electrical equipment, By applying a signal separation technique based on estimation means such as independent component analysis, it is possible to connect from the result of measuring the total load current at one point of the power customer's power line, for example, the power line inlet to which the monitored device is connected. In addition, it was considered to estimate the individual operating conditions and current consumption of each electrical device.
[0011]
That is, the invention according to claim 1 is obtained from a measurement sensor installed on a power consumer's power supply line in a remote monitoring system that estimates individual current consumption of a plurality of electrical devices used by a power consumer. The effective current of each harmonic of the total current is used as an input, and based on the signal separation algorithm, the consumption current for each device of the plurality of electric devices used by the power consumer is provided as an estimation means. That is, the remote electrical equipment monitoring method of the present invention is a remote electrical equipment monitoring method for estimating the movement status of a plurality of electrical equipment used by a power consumer, and measures the total load current from the power consumer's feeder line. The total load current is converted into current for each fundamental wave and harmonics, and current change data is created by taking a time difference between the currents for each fundamental wave and harmonics. Current component is estimated for each component estimated as a device group having the same harmonic intensity ratio by independent component analysis, and the current consumption waveform for each component of the same harmonic intensity ratio component It is characterized by estimating a current change.
[0012]
Moreover, this remote electrical equipment monitoring method is implement | achieved by the remote electrical equipment monitoring apparatus which estimates the movement condition of the several electrical equipment currently used by the electric power consumer concerning the invention of Claim 4, for example. The remote electrical equipment monitoring device includes a total current sensor that measures a total load current from a power consumer's power line, a frequency component conversion device that converts the total load current into a fundamental wave and a harmonic current of the total load current, and A time difference device that obtains a current change by taking a time difference between the fundamental wave of the total load current and the current for each harmonic, and the current change is estimated as an equipment group having the same harmonic intensity ratio by independent component analysis An independent component analyzer that separates each component, and a device-specific signal separator that estimates the movement status (current change) of each device to be monitored from the waveform of the current change for each component of the same harmonic intensity ratio ing.
[0013]
A device with built-in inverter circuit and rectifier circuit generates unique harmonics. The harmonic effective current itself at each time of each device may be generally similar, for example, between devices having the same start time, such as increasing with time. However, the change in the effective current (time differentiation) of each harmonic is highly statistically independent even if the devices started at the same time are not activated unless a special synchronization mechanism is activated (this is the individual device current change). Called statistical independence). In addition, the intensity ratio of the effective current (change) of each harmonic of each device is substantially stable over time. In addition, the intensity ratio is unique in an individual device or a device group having similar harmonic characteristics (for example, a device group that does not emit harmonics). For this reason, it is possible to estimate the current change of each individual device by the weighted sum of the effective current changes of the respective harmonics of the total current (this is called linearity of the individual device current change). The harmonic ratio specific to each device group and the weight of each harmonic of the total current are in an inverse matrix relationship.
[0014]
Based on the linearity and independence of the individual device current changes described above, by estimating the weight that maximizes the statistical independence between the individual current changes of each device estimated from the harmonics of the total current, It is possible to estimate the current change of each device group having a specific harmonic intensity ratio.
[0015]
Further, in the invention according to claim 1 or 4, when current changes of a device group having a specific harmonic intensity ratio are separated into one, based on an operation characteristic model relating to the intensity level of the current change of each individual device. Thus, it is possible to separate individual current changes of each device having the same harmonic intensity ratio. That is, the invention according to claim 2 is the remote electrical device monitoring method according to claim 1, wherein the component of the device exhibiting the same harmonic intensity ratio among the current change for each component of the same harmonic intensity ratio is determined by the monitored device. Based on the information on the current change intensity, the current change of individual devices is estimated by further separation. Further, the invention according to claim 5 is the remote electrical equipment monitoring apparatus according to claim 4, wherein the component of the equipment exhibiting the same harmonic intensity ratio among the current change for each same harmonic intensity ratio component is the Based on the information on the current change intensity, the current change of individual devices is estimated by further separation.
[0016]
Further, the remote electrical equipment monitoring method according to claim 1 or 2 and the remote electrical equipment monitoring device according to claim 4 or 5 are separated for each component estimated as a device group showing the same harmonic intensity ratio output. Since the current change is device-specific, the power consumption can be estimated by obtaining the current consumption from these current changes as in the inventions of claims 3 to 6.
[0017]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, the configuration of the present invention will be described in detail based on an example of an embodiment shown in the drawings.
[0018]
In FIG. 1, the outline | summary of one Embodiment of the remote electrical equipment monitoring method of this invention is shown. This remote electrical equipment monitoring method is for estimating the movement status of a plurality of electrical equipment used by a power consumer, measuring the total load current from the power consumer's power supply line, and calculating the total load current. The current is converted to current for each fundamental wave and harmonic, and current change data is created by taking the time difference between the current for each fundamental wave and harmonic, and independent component analysis is performed for the current change for each fundamental wave and harmonic. To separate each component estimated as a device group having the same harmonic intensity ratio, and output the current change for each device of the monitored device from the current change waveform for each component of the same harmonic intensity ratio. The operating state can be estimated.
[0019]
Here, separating and estimating the current consumption of individual devices from only the total current consumption, more generally, when signals from a plurality of signal sources are added together and measured as one signal at one location, It is considered that this is one of the methods of separating the signals of each signal source from the measurement signal and tracking the change, that is, “signal separation tracking”. In this case, the “signal source” corresponds to the device, the “signal measured at one place” corresponds to the total current, and the “signal of each signal source” corresponds to the “current (current change) of each device”.
[0020]
First, measure the total current, which is the sum of the power consumption of each device, at a high sample rate, perform Fourier transform on the measured total current, and calculate the effective frequency (fundamental harmonic of the total load current). Convert to time series of current values. Then, signal separation based on independence is performed. The current of each device changes autonomously unique to each device. In other words, it fluctuates independently of current changes in other devices. Further, the operation of each device changes simultaneously in a plurality of frequency bands, but in which frequency band a strong change is made differs depending on the device. Therefore, by analyzing independent current changes that occur simultaneously in a plurality of frequency bands, it is possible in principle to restore the current consumption of each device. Due to this property, it is possible to restore the current consumption of each device without an operation model unique to the device.
[0021]
However, if the rated current of the monitored device is known in advance, or if the total current is continuously monitored and the device current is restored appropriately, a device-specific operating characteristic model is configured. be able to. The operating characteristic model can be expressed by a probability distribution of possible values of the amount of change in the effective current I (t) of the device (I (t + 1) -I (t)) (hereinafter referred to as current change). .
[0022]
In the present invention, when a device-specific operation characteristic model can be used, it is used to improve estimation accuracy. When using a device-specific operation characteristic model, the use of an operation characteristic model that matches the behavior characteristic of the device in operation improves the separation accuracy. For this reason, in order to continuously track the operation of the device, it is necessary to select the device operation characteristic model used in step 2 of FIG. In the present embodiment, on the assumption that the operations of a large number of devices do not change at the same time, a device operation characteristic model to be used is selected, and the process returns to Step 2.
[0023]
The above is an overview of the signal tracking approach. The following outlines the algorithm used in each step.
[0024]
[Signal separation based on independence]
The signal separation method is based on the standard independent component analysis method and is expanded to allow separation that takes into account the operating characteristics of the equipment.
[0025]
The independent component analysis method is a method of estimating a signal waveform of a signal source from an observation signal, for example, separating sounds emitted from each sound source from simultaneous recording at a plurality of points. In the application, the following two assumptions of “linearity” and “independence” need to be established.
[0026]
“Linearity Hypothesis 1”
(1) The k-th harmonic current change amount dIi (t) of the total current is expressed as the sum of the k-th harmonic current change amounts dSk, i (t) of each device.
dIk (t) = Σj dIk, j (t)
(Strictly established when the current value and current change are displayed in complex numbers)
(2) The k-th harmonic current change amount dSk, i (t) of the device i is
Device i's normalized harmonic current variation dSi * (t) is a fixed multiple Ak, i
dSk, i (t) = Ak, i ・ dS * i (t)
DIk (t) = Σj Ak, i ・ dS * i (t)
In matrix representation, dI = A ・ dS *
Where dI = (dIk (t)), A = (Ak, i), dS * = (dS * i (t))
Represented as: That is, “linearity hypothesis 1” is equivalent to the following “linearity hypothesis 2” when an inverse matrix of A exists.
[0027]
“Linearity Hypothesis 2”
(1) Normalized harmonic current change dS * i (t) of device i is
This is the sum of Bk, i, which is a fixed multiple of each harmonic current change dIk (t) of the total current.
dSi * (t) = Σk Bi, k ・ dIk (t)
In matrix representation, dS * ＝ B ・ dI
Where dI = (dIk (t)), B = (B, i, k), dS * = (dS * i (t))
(2) The amount of change in the k-th harmonic current dSk, i (t) of device i is
Device i's normalized harmonic current variation dSi * (t) is a fixed multiple Ak, i
dSk, i (t) = Ak, i ・ dS * i (t)
Where Ak, i is the (k, i) component of the inverse matrix of B. The matrix A is called “mixing matrix”, and the matrix B is called “separation matrix”.
[0028]
Under the linearity hypothesis, if the separation matrix B is known, the specified harmonic current change amount of the device can be estimated from the harmonic current change amount of the total current. However, devices having the same frequency intensity Ak, i ratio cannot be distinguished because there is no inverse matrix. Therefore, strictly speaking, a device group having the same frequency intensity ratio is virtually treated as one device.
[0029]
"Independence"
The current waveforms of each signal source device are independent of each other. That is, the independence of the current change amount of the device is derived from the fact that the operation of each device (switch on / off or operation mode change) is performed independently of the operation of other devices.
[0030]
Mathematically, the current change amount dS of device i, j^* _i(t), dS^* _jThe independence of (t) is
・ Current change amount dS of device i at time t^* _iProbability Pr (dS for (t) = x^* _i= x),
・ Current change dS of device j at time t^* _jProbability Pr (dS for (t) = y^* _j= y),
・ Current change amount dS of device i at time t^* _i(t) = x and device j current change dS^* _j(t) = y
Probability Pr (dS^* _i= x, dS^* _j= y)
When Pr (dS^* _i= x, dS^*j = y) = Pr (dS^* _i= x) ・ Pr (dS^* _j= Y) is established. At this time, the estimated signal of each signal source is called an independent component.
[0031]
This standard independent component analysis does not have a source-specific characteristic model and assumes that the signal independence of each source is only assumed that the observed signal is a linear superposition of signals from independent sources. The signal source S is estimated by obtaining the separation matrix B that maximizes the index (independence index) indicating That is, in the present invention, even when the separation matrix B in the linearity hypothesis 2 is unknown, focusing on the independence of the current change amount of the device, applying the independent component analysis method, the estimated harmonic of the device. The separation matrix B that maximizes the independence index between the current amounts is obtained. Furthermore, according to the linearity hypothesis 2 (2), an inverse matrix A of the separation matrix B, which is a frequency intensity matrix, is obtained, thereby estimating a fundamental current change amount for each device group having the same frequency intensity ratio.
[0032]
Thereby, it is possible to estimate the basic current amount change for each device group having the same frequency intensity ratio from the harmonic current change of the total current.
[Improvement of accuracy by equipment operating characteristic model]
Since the standard independent component analysis described above does not have a characteristic model specific to the signal source and the characteristic of the signal is completely unknown, a signal source having a similar characteristic may cause interference. Therefore, if the ratio of the current change amount by device and the rating of the current change of the device are known, the basic current change amount obtained by the above standard independent component analysis is calculated based on that information. By separating into changes, it is possible to estimate the amount of current change for each device.
[0033]
In the proposed method, an operation characteristic model unique to each signal source (device) is obtained as a possible signal level S._iProbability distribution Pr of (t)_iThis is given as (s) and is used to improve separation accuracy.
[0034]
The change in current consumption of the device tends to be extremely concentrated around a specific value. For example, in an on / off type device, it is either 0 or a rated value. The same tendency occurs in inverter equipment.
[0035]
Two measures are taken to improve the separation accuracy. First, the independence index that is maximized by the independent component analysis is optimized based on the behavior characteristic model. By maximizing the obtained independence index, signal separation suitable for the device characteristics is performed.
[0036]
Second, the device-specific operation characteristic model is used for the separated independent components to separate the corresponding device signals. This is effective for signal separation when interference occurs because the operation characteristics except for the signal level are similar.
[0037]
FIG. 2 shows an embodiment of an apparatus for realizing the remote electrical equipment monitoring method of the present invention. This remote electrical equipment monitoring apparatus basically includes a total current sensor 11 that measures the total load current from the power supply line 2 of the power consumer 1, and a fundamental wave and a harmonic current of the total load current from the total load current. A frequency component conversion device 12 for converting the current to the current, a time difference device 13 for obtaining a current change by taking a time difference between the fundamental wave of the total load current and the current for each harmonic, and the same harmonic by the independent component analysis. The independent component analyzer 14 that separates each component estimated as a device group having an intensity ratio, and the movement status (current change) of each device to be monitored 3 from the waveform of the current change for each same harmonic intensity ratio component. And a device-specific signal separation device 15 to be estimated.
[0038]
The measurement sensor 11 is installed only at one location in the vicinity of the service port 2 of the service customer 1 in order to make the system non-intrusive. The measurement sensor 11 obtains an electric current and is constituted by a current transformer, for example. In this embodiment, the case where it is implemented as an example for a general electric power consumer in Japan that uses a single-phase three-wire lead-in wire is taken as an example. Therefore, a current transformer for an A phase and an instrument for a B phase It consists of a current transformer. For example, assuming that a through-type current transformer is used for the A-phase and B-phase instrument current transformers, the instrument current transformer measures the current flowing in the A-phase on the primary side and starts the A-phase from the secondary side. The current transformer for measuring the current flowing in the B phase on the primary side and outputs a current similar to the B phase current from the secondary side. These currents are input to the fast Fourier transform device 12 which is a frequency component conversion device.
[0039]
The fast Fourier transform device 12 extracts data relating to the fundamental wave of the total load current and the current for each harmonic from the total load current detected by the measurement sensor 11. Specifically, although not shown, for example, an analog / digital (A / D) converter and a Fast Fourier Transformer are used, and A-phase and B-phase currents I input from the measurement sensor 11._A, I_BIs converted into digital data by an A / D converter, and then harmonic current data I is converted by a fast Fourier transformer._{A (1-13)}, I_{B (1-13)}Have been trying to get. Here, the current data I_A1, I_B1Indicates the fundamental current of the total load current, and the current data I_{A (2-13)}, I_{B (2-13)}The suffix (2-13) indicates the harmonic order, that is, the harmonic current representing the second to thirteenth orders. The fundamental frequency of the AC power supplied to the feeder is multiplied by the numerical value of the order. Represents the frequency of a wave. For example, when the fundamental frequency is 50 Hz, the third harmonic current refers to a current component having only a frequency component of 150 Hz. In general, odd-order harmonics appear predominantly and even-order harmonics are small, and therefore, the fundamental wave and odd-order harmonic data are given to the time difference device 13 as inputs here.
[0040]
The time difference device 13 obtains and outputs a current-current change amount by taking a time difference between the fundamental wave and the harmonics of the total load current converted by the Fourier transformer 12.
[0041]
This independent component analyzer 14 separates current changes for each component estimated as a device group having the same harmonic intensity ratio by independent component analysis, and is independent by a computer that executes the algorithms of FIGS. The component analysis is performed. Further, the device-specific signal separation device 15 estimates the movable state (current change) for each device of the monitored device 3 from the waveform of the current change for each same harmonic intensity ratio component, and executes the algorithm of FIG. Independent component analysis is performed by a computer.
[0042]
3 to 6, the fundamental current change amount for each device group having the same frequency intensity ratio is estimated by a standard algorithm and an algorithm using an operation characteristic model.
[0043]
First, the current value dIk (t) = Ik (t) −Ik (t−1) of the odd-order harmonics from the first order (fundamental wave) to the 13th order (harmonic) of the total current is input (step S1). ).
[0044]
Then the separation matrix B = (B_{i, k} ) Is set (step S2).
The settings are as follows.
U is a random n-th order rotation matrix,
When the eigenvalue decomposition of the covariance matrix of the first to nth harmonic current change amount dIk (t) of the total current is eigenvalue decomposition V · D · Vt, B = U · D−1 / 2 · V.
The covariance matrix is the covariance Σ_t (dI_k(t) -dI_kAverage) ・ (dI_l(t) -dI_lThe average) / T.
[0045]
Next, normalized current change by device group dS *_i(t) = Σ_iB_{i, k}・ DI_k(t) is estimated (step S3).
[0046]
Next, the independence index IND (dS^*) Is improved by the natural gradient method or the fixed point independent component analysis (step S4).
Here, various independent component analysis methods can be used as the independent component analysis method used in the present invention. Current change dS^*i, dS^*Independence index IND (dS^*)
・ Σ_i(Σ_t dS^* _i(t)^Four / T-3)²       : Sum of squares of kurtosis of each device
・ Σ_i(Σ_t log (cosh (πdS^* _i)) / T)²
The sum of squares is used.
[0047]
Here, it is effective to use the sum of kurtosis as an index in order to improve the estimation accuracy of a large current change amount of the device such as an on / off operation. Further, in the present invention, when there is already a current record of the monitored device, the independence IND (dS) with high separation accuracy for these waveforms is determined by the algorithm shown in FIG. Increase the separation accuracy of monitored devices.
[0048]
Algorithms that maximize the independence index include:
・ Fixed point independent component analysis (Fixed Point ICA) to find the fixed point of IND (dS)
・ Fast JADE algorithm when IND (dS) is Kurtosis sum of squares
・ Improved IND (dS) mountain climbing method toward natural gradient (especially suitable for online update)
There is.
[0049]
In the present invention, when the independence index IND (dS) is obtained by the algorithm X, updating by the natural gradient method (separation matrix updating algorithm in FIG. 5) is used.
[0050]
Next, if the number of updates of the separation matrix B is greater than or equal to the set value, or if the update amount is greater than or equal to a certain value, the process jumps to step 4; otherwise, the process jumps to step 6 (step S5).
[0051]
Next, the estimation of the mixing matrix A is set to an inverse matrix of A = B (step S6). A_{k, i}Is the harmonic intensity of each device group.
[0052]
Next, the fundamental current change amount dS of each device group_ian estimate of (t), that is
dS_i(t) = A_{1, i} ・ DS^* _i(t) is executed (step S7).
[0053]
Then, the fundamental current change amount dS of each device group i_i(t), fundamental wave and each harmonic intensity A_{k, i} Is output (step S8).
[0054]
The estimation of the current change amount for each device from the device group basic current change amount obtained in step 8 is performed according to the algorithm of FIG. 6, for example.
That is, the amount of current change by device group dSi (t)
Device group k-th frequency intensity Aki
Frequency intensity Ckj during operation of the target device j
(In the case of an on / off type device, the fundamental wave Ckj is the rated current, and the higher order Ckj = 0)
Is input (step 9), and initial setting dS ′ (t) = 0, t = 1..T of the current change amount estimation of the device j is performed (step 10).
[0055]
Next, it is determined whether the frequency intensity ratio matches (step 11). This is the cosine (cos) = | Σ of the angle between the frequency intensity vector {Aki, k = 1..n} of the device group i and the frequency intensity vector {Ckj, k = 1, .. n} of the device j._k A_ki・ C_kj| / √ ((Σ_kA_ki2) (Σ_kC_kj2)) is a constant value 1-ε₀If it is below, the process ends as nonconforming.
[0056]
Next, the current change amount matching determination is performed (step 12). This is because, for time t when dSi (t) of device group i is within fundamental frequency intensity C1j ± ε of device j, the current change is determined to be due to device j.
dS’j (t) = dSi (t), dSi (t) = 0
And
[0057]
Then, the current change dS′j (t), t = 1..T of the device j and the current change dSi (t), t = 1..T that cannot be explained by the operation of the device j are output (step 13).
[0058]
The optimum independent index is estimated according to the algorithm shown in FIG. 4, and the update of the separation matrix B in step 4 is performed according to the separation matrix update algorithm when the optimum independent index G shown in FIG. 5 is used.
[0059]
According to the remote electrical equipment monitoring apparatus configured as described above, unknown measurement data (current I from the measurement sensor 11 installed in the vicinity of the feeder inlet._A, I_B ) Is taken out from the fast Fourier transform device 12 and separated by the time difference device 13, the independent component analysis device 14, and the device-specific signal separation device 15 into current changes for each electrical device group having the same frequency characteristic, that is, the same ratio, and output. The Therefore, it is possible to estimate the individual operating status and current consumption (and thus power) of the electric equipment 3 (non-inverter equipment, inverter equipment) from the waveform of this current change.
[0060]
In order to confirm the usefulness of the present invention, the following experiment was conducted.
Assume that an ammeter is connected to the power line at the service entrance and the total current used is measured in seconds. The effectiveness of the proposed method is verified by estimating the current used per second for each device, which is a breakdown from this measurement signal. For this purpose, data on the current used and total current of each device per second is required. The data comprises an ammeter that measures the current used for each electrical device that is a load and an ammeter that measures the total load current, and consists of a fast Fourier transform device that converts the total load current into a fundamental wave and a harmonic. Obtained by a measuring device (not shown). Here, the measuring device includes a sine wave power supply device (50 Hz, 100 V, 2 KVA) and a switch for switching the operation state for each individual load. For example, in the case of an inverter air conditioner, the indoor set temperature and the set wind speed are used. By changing the current, the current of the non-inverter device (for example, if it is an incandescent lamp, the current can be changed by increasing / decreasing the number of lights) is set arbitrarily. Various combinations of load usage conditions can be obtained.
[0061]
The current value per second of each device is measured using an ammeter connected to each device. The measured value of each ammeter is the effective current value of the fundamental wave (50 Hz) component. On the other hand, strictly speaking, the total current is not the total current per second. The total current is measured once every second for 5 fundamental waves (5/50 seconds), and the high frequency of the odd order from the fundamental wave (50Hz) to the 13th order (650Hz) obtained by the fast Fourier transform (FFT). The effective current value, effective voltage, and phase angle of voltage and current are recorded. In the analysis, voltage and phase angle are not used, and only the effective current is focused.
[0062]
Since the measurement method is different, the total current value does not always match the total current value of the ammeter of the device. The total error between the total current value and the current value of each device was an average of -0.1A and a standard deviation of 0.05A. The rate of error of -0.1 ± 0.4 A or more was 0.01% or less.
[0063]
In the experiment, 15 devices shown in Table 1 were connected and measured.
[Table 1]

Among them, 256 (= 2) which changed ON / OFF of the switch of 8 devices (refer to the last column of Table 1)⁸) The case is measured for about 5 minutes (900 seconds). Among the other seven models, two devices, the pot and the refrigerator, entered the heat insulation / cooling mode autonomously by automatic operation. However, since five devices such as the microwave oven were connected in the off state, there was almost no current consumption. The incandescent lamp has five bulbs of the same type, but in the case where this is turned on, the lamps were turned on in order of 1, 2,. The same operation is performed for fluorescent lamps and inverter fluorescent lamps.
[Features of electrical equipment waveforms and current hypothesis of independence]
In order to perform signal separation based on the assumption of independence, the waveforms that result from the separation need to be independent of each other, that is, change without relation. However, this hypothesis does not necessarily hold for the current waveform of electrical equipment.
[0064]
As an example, FIG. 7 shows a case 197 (11000101) in which two air conditioners (A1, A2), an inverter type fluorescent lamp (F1), and one television (T3) are turned on. The upper part of FIG. 7 is a graph of the effective total current value of each device for 5 minutes (900 seconds).
[0065]
It can be seen that the air conditioners A1 and A2 are changing in a similar manner. This is a result of starting both air conditioners almost simultaneously. For this reason, if it isolate | separates based on the hypothesis of independence, the component of a common behavior and difference component of both air-conditioners will become each independent component. Since it is often seen that multiple devices are switched on at approximately the same time, it is difficult to directly assume independence between the current values of the electrical devices.
[0066]
On the other hand, the lower part of FIG. 7 shows the change amount of the effective current value of each device at each time (the difference between current values at time t and time t + 1, hereinafter, current change (amount)). It can be seen that there is little similarity to the time change of the current change amount of each device and the independence is high. Even when the overall trend of device operation is similar, the timing when each device enters a specific operation mode, such as when the air conditioner enters cooling operation (rising edge of effective current value), unless there is a special synchronization mechanism, It is controlled autonomously by each device. For this reason, it can be estimated that the independence of the current change is high.
[0067]
The independence of the current change of each device reflects the autonomy of the operation of each device, and can be expected to be an effective hypothesis that generally holds. Note that the separation matrix obtained from the current change can be used as it is as the separation matrix for the effective current.
[0068]
In order to verify the independence hypothesis, the standard independent component analysis algorithm JADE (JF Cardoso et. Al .: “Blind beamforcing for non-Gaussian signals”, IEE Proceedings -F, 140 ( 6): 362-370, 1993.) and a case where the same algorithm was applied to the current change of the operating equipment were compared for all 256 cases. The result is shown in FIG. In this experiment, since the effective current and current change of each operating device are normalized to variance 1 and given as an input signal, the input signal itself is ideally an independent component. That is, the separation matrix should be a unit matrix. FIG. 8 shows the deviation between the obtained separation matrix and unit matrix (difference between the minimum absolute value of the diagonal component and 1). 0 means that complete separation / reproduction is possible, and the closer to 1, the better the separation is and the interference is.
[0069]
In the case of current change, the degree of interference is 0.57 at the maximum, and in 217 cases out of 256 cases, high-precision separation with a degree of interference of 0.05 or less is achieved, whereas in the case of effective current, Has occurred in 132 cases with 0.9 or more interference.
[0070]
It was confirmed that the independence hypothesis of the current change of each device is effective in separating the effective current of the device.
[0071]
[Effective current by device and linearity hypothesis]
Next, the relationship between the effective current of each frequency of the measured total current and the effective current of each device to be estimated will be examined.
[0072]
FIG. 9 is a time change of each frequency component and current change (amount) of the air conditioner (A1, A2) in case 197. When attention is paid to the point in time when the air conditioners A1 and A2 are switched from the cooling mode to the steady operation (the dotted line portion in the vicinity of time 150 seconds in FIG. 9), the third wave is generated in the same manner, whereas It can be observed that the next wave is in the opposite direction and that only the air conditioner A2 is strongly influenced in the 13th wave. Thus, the change in the operation mode of each device contributes differently to the current change value of each frequency component. Therefore, there is a possibility that the current change of each device can be estimated from the current change of each frequency component. The problem here is whether or not the assumption of “linearity” required in independent component analysis, that is, the property that the waveform of each signal source is expressed by linear synthesis of the waveform of each observation signal, is established. It is.
[0073]
Since the sum of the currents from each device is the total current, this hypothesis is that the ratio of the effective current (Ij) at each frequency of the device i is device-specific and is constant regardless of the output level and time.
(I₁(t), I_Three(t), ..., I₁₃(t)) = (a₁,…, A₁₃・ I₀(t)
Means that This hypothesis that “the ratio of the effective current at each frequency of the device does not change with time and has a value peculiar to the device” does not hold strictly like the independence hypothesis. However, if established approximately, the outline of the current change of the device current can be estimated from the effective current (current change) of each frequency component of the total current.
[0074]
In the experiment, the effective current at the fundamental frequency is measured for each device alone, but the effective current for each frequency is not measured, so it is difficult to directly verify the above hypothesis. Here, the assumption of the hypothesis is verified by examining the prediction accuracy of the effective current at the fundamental frequency of the device when it is assumed that the hypothesis is established using the effective current for each frequency of the total current.
[0075]
FIG. 10 shows the result of obtaining the separation matrix B from the effective current value of each frequency component and the effective current value of the device by minimizing the mean square error (least square method). The upper part of FIG. 10 is an approximate effective current value of each device obtained by the separation matrix B from the effective current value at each frequency. The lower part of FIG. 10 is a difference value (hereinafter referred to as an approximate difference current) of approximate effective current values of each device. Compared with FIG. 7, for the air conditioners (A1, A2) and the like, it can be seen that the approximate effective current is close to the actual effective current, and the linearity is approximately established. On the other hand, for the inverter fluorescent lamp (F1) and the pot (J), the approximate waveform is very different from the actual effective current, and the assumption of linearity does not give a good approximation. However, with respect to the total value (F1 + J) of both, the approximation accuracy is improved rather than estimating each independently, and it is considered that linearity is approximately established. As estimated from the waveform similarities of the approximate difference current values (FIG. 10), the inverter fluorescent lamp and the pot have similar frequency characteristics (ratio of effective currents at each frequency (a1 / a1,... This is probably because of having a13 / a1)).
[0076]
Therefore, the assumption of linearity does not necessarily hold for individual devices, but is approximately established if a specific device group having similar frequency characteristics is regarded as one device. In order to further separate devices with similar frequency characteristics, such as inverter fluorescent lamps and pots, the intensity level I of the fundamental wave of each device₁＝ a1 ・ I₀It is necessary to introduce a hypothesis about the above, that is, the value of the current change specific to the device (eg, pot J: 0. 6A). (As will be described later, this may be achieved by using a device operating characteristic model.)
[0077]
In order to examine a device group based on frequency characteristics, obtain the correlation coefficient of the approximate differential current between the operating devices in each case and the approximate error of the approximate differential current (approximate differential current-actual current change), and calculate the absolute value. The averaged values are shown in FIG. As discussed in the previous section, the actual current change itself is highly independent and does not show a strong correlation. A large correlation coefficient of the approximate difference current means that the devices have similar frequency intensity ratios, and it is predicted that the current changes of both devices are easily separated as one independent signal component. On the other hand, the correlation coefficient of the approximate current change error indicates the similarity in the behavior of the portion of the current change that cannot be captured by linearity. Therefore, although a device with this high correlation is separated as separate independent signal components, it is predicted that interference between the two tends to occur at a specific time.
[0078]
From FIG. 12, the device group {incandescent lamp (4), fluorescent lamp (5), inverter type fluorescent lamp (6), pot (9)} has a high correlation coefficient (linear part) of the approximate differential current, and is independent. It can be predicted that the component analysis is difficult to distinguish in the component analysis. In addition, the device group {air conditioner A2 (2), electric fan (3), television 1 (7)} is also highly similar in terms of linearity. It can be seen that the air conditioners (A1 (1), 2 (2)) have non-linear similarities between the electric fan (3) and all devices other than the television 1 (7). In particular, air conditioners are strongly related to each other, and partial interference is likely to occur.
[0079]
[Estimation of current change by device]
[Application experiment of standard algorithm]
It has been confirmed that independence with respect to current change by device and linearity with respect to an appropriate device group can be assumed. This section describes an experiment that estimates the current change per second of an unknown connected device from only the effective current value per second of each frequency component, based on the above assumption of independence and linearity.
[0080]
In the experiment, first, the standard independent component analysis method JADE was applied to estimate the current change of individual devices. In device operation estimation, spike-like current changes such as on / off, which can be called a kind of exceptional value, are important, but JADE efficiently maximizes the sum of squares of the independence index kurtosis, which is strongly influenced by exceptional values. To do. For this reason, the JADE algorithm was selected.
[0081]
FIG. 12 shows the result of estimating the waveform (current change in the fundamental frequency) of the signal source with the seven frequencies (first to seventh stages in FIG. 9) from the fundamental wave to the 13th order (odd order) in case 197 as input. . j-order harmonic current change X_j, Mixing matrix A, separation matrix B, the current change of the fundamental wave of the i-th estimated component is S_i= Σ_jA (1, i) ・ B (i, j) ・ X_jIt is obtained by. The first component (maximum amplitude 3A), the second component (2.8A), the third component (0.8A), and the fourth component (0.4A) are the main components. FIG. 13 shows the first to third components and the waveform of the corresponding operating device (current change of the fundamental wave).
[0082]
The first component reproduces the air conditioner A1 and the second component reproduces the air conditioner A2 relatively well. However, if it sees in detail, the component at the time of starting of air-conditioner A2 will have some interference in the 1st component (300 seconds and 600 seconds vicinity). When the mixing matrix A is examined, the first and second component frequency intensities are fundamental waves, third-order waves, and fifth-order waves, the first component is about 10: 6: 2, and the second component is 10: 2. The intensity ratio is 5: -1.
[0083]
The third component does not correspond to a single device. Spike component corresponding to lighting of inverter fluorescent lamp (4 times, approx. 0.2A) and start / stop of pot (5 times, approx. ± 0.7A) and part of air conditioner A2 (small spike ( 3 times, -0.5A) and part of start-up) are mixed. Although not shown, from the experimental results, the fourth component (two spikes, 0.4 A) corresponds to the second and third start-up peaks of the air conditioner A2. The fifth component has a strong correlation with the television.
[0084]
In the third component, the intensity ratio of the fundamental wave and the tertiary wave is 10: 1, and the intensity of other frequencies can be ignored. In other words, it is an equipment component with almost no harmonics. The fact that the pot and the inverter fluorescent lamp are included in the third component is consistent with the discussion of the similarity between the inverter fluorescent lamp F1 and the pot J described above. The reason why the air conditioner temporarily interferes with the third component is thought to be because the harmonics were temporarily weakened by the inverter control of the air conditioner. Regarding the air conditioner A2, the waveform at the start-up particularly interferes with other components other than the corresponding second component, and some spikes are separated as the fourth component. This is considered to be caused by nonlinearity due to inverter control of the air conditioner. This point is also consistent with the above discussion (see FIG. 11B).
[0085]
The above experiment showed that there is some interference due to the same frequency intensity characteristics and non-linearity, but from the effective current value of the odd frequency component of the total current, it is possible to almost reproduce the operating status of each major device. .
[Accuracy improvement by equipment-specific model]
In general independent component analysis, it is assumed that the characteristics of the separation target device are completely unknown, and each signal source s_iAssuming the same independence index H (s) regardless of the nature of_iH (s_i The separation matrix is determined so that) is minimized.
[0086]
However, in the signal separation tracking problem in which regular monitoring is performed, particularly in the case of device operation signal separation, the type of connected device can be assumed to some extent in advance, and the characteristics of the electric device can be grasped in advance from its rating. Even if it is unknown in advance, it may be identified as a signal source having the same characteristics as signal separation is constantly advanced. In such a case, the separation accuracy can be improved by utilizing the operation characteristics unique to the device.
[0087]
In the proposed method, the device operation characteristic is given by the standard current change value of the current change (fundamental wave) of the existing device. For example, in an on / off type device, the current change value during the on operation and the current change value during the off operation take substantially constant values.
[0088]
In the proposed method, when these standard distributions Qi (x) are given, an appropriate independence index H (s) = Σ is obtained by independent component analysis based on signal source characteristics, for example, independent component analysis including the algorithm of FIG._iH (s_i) And maximize it.
[0089]
FIG. 15 shows the optimum independence index (main factor) obtained when a standard probability distribution of an air conditioner or the like is given and the independence index used in the representative method. JADE shows good separation performance compared to other independence indices, as compared to the independence indices used by JADE (x^Four) Is a good approximation of the optimal independence index.
[0090]
In the proposed algorithm, the interference generated in the separated independent components is further eliminated based on the given device operating characteristic model.
[0091]
In the example of Case 197, the third component shows some interference from the inverter fluorescent lamp (F1), pot (J), and air conditioner (A2). (s) (Fluorescent lamp: Mixture distribution with two peaks at 0A in steady state and about 0.2A when lit, Pot: Mixture with peak at about ± 0.65A at steady state 0A and start / stop Distribution).
[0092]
FIG. 16 shows the device current change finally estimated using the device-specific model and the extended independence analysis algorithm. It can be seen that the separation accuracy of devices with few harmonics has been improved by incorporating device-specific operation models.
[0093]
From the above, it was shown that the proposed method focusing on the independence of the proposed device operation is effective for device operation separation tracking. Furthermore, by applying the proposed method to data measured by 10 operating home appliances, there is no need to install a measurement device or information collection device in individual devices, and information on the operating characteristics of connected devices can be obtained. It was shown that the operating state of major equipment can be estimated without it.
[0094]
In addition, when a device operating characteristic model was given, it was shown that the estimation accuracy can be improved by recognizing a current change recognized as a current change of one device due to interference as a current change of a different device.
[0095]
In the device operation signal separation problem, the component corresponding to the current change (current change) of the main connected device is separated by paying attention to the fact that the property of the independent operation of the device appears in the current change (current change) of the device. I was able to. This property is based on the very general property that devices operate asynchronously and autonomously.
[0096]
However, the linearity hypothesis, that is, the property that the intensity ratio of harmonics of each device is unique to the device and almost constant in time, does not strictly hold. There are two causes for this. For one thing, there are devices with similar intensity ratios such as inverter fluorescent lamps and pots, and non-linearity that the intensity ratio varies from time to time as in air conditioners. The first cause causes identification of a plurality of devices having similar intensity ratios, and the second cause causes signal dispersion such as interference with other components and separation of a single device into a plurality of components. These cause a decrease in estimation accuracy.
[0097]
For devices with similar intensity ratios, it is shown that if a device-specific operating characteristic model is given, it can be recognized as a current change of another device by the proposed method based on the strength information obtained from it. The dispersion of the signal due to the nonlinearity is not so severe that the corresponding signal component (the third component in the air conditioner 2 of case 197) becomes unknown, but remains at an effective level for estimating the operation state. However, in order to improve the estimation accuracy and estimate the operating state of the small current device, a countermeasure is required.
[0098]
Focusing on the independence, the technology for estimating the operation of the connected device is a very useful technology in that the operation can be estimated to some extent without preparing a device-specific model in advance.
[0099]
The above-described embodiment is an example of a preferred embodiment of the present invention, but is not limited thereto, and various modifications can be made without departing from the scope of the present invention. For example, in this embodiment, the effective current value of the fundamental wave of each device, that is, the effective current value I₀Although the method using the time difference δI (t) of (t) is described, the current value of the harmonic of each order is determined by the phase difference ω (t) of the current and voltage of the harmonic order of each order.
I (t) = √2 · X₀(t) ・ (cos (ω (t)) + i ・ sin (ω (t)))
As a complex expression (i: imaginary number), it can be easily extended to currents including reactive currents.
[0100]
By using the independent component analysis method for the complex matrix for the time difference δI (f, t) of the harmonic current value (complex expression) of each order of the total current value, similarly, the fundamental wave of each device The time difference ΔS (i, t) of the current value (complex expression) can be estimated. This method corresponds to a case where the phase difference ω (t) is approximated to be substantially constant.
[0101]
In the present embodiment, the estimation of individual power consumption of a non-intrusive power device has been described. However, the usage method is not particularly limited, and it can be used to warn of abnormal operation of an electrical device. That is, when it is determined from the information on the power consumption obtained by the power consumption estimation system of the electrical device that the abnormality is present in, for example, comparison with daily power consumption, the safety of the power consumer occupants and the safety within the power consumer It is possible to determine the presence / absence of an abnormality in an electric device or an electrification system and to transmit the information to the outside. For example, with this system, it is possible to determine the “safety of occupants of electric power consumers” from the operating state of lighting, television, electric kettle, hot water toilet seat, etc. that the occupants should be turned on and off every day, fire, etc. It is possible to determine “safety in electric power consumers” from long-term use (keep on) such as an electric iron, an electric stove, and an electrified kitchen that cause the above. With regard to “external transmission of these information”, existing telephone lines, PHS, pagers, the Internet, etc. can be used, and “reported persons” are those who are resident, those who are resident, fire departments, local governments, etc. We can assume welfare medical person in charge.
[0102]
【The invention's effect】
As is clear from the above description, according to the remote electrical device monitoring method and apparatus of the present invention as set forth in claims 1 and 4, the current change of one device is independent of the current change of other devices. Focus on and use the independent component analysis method to separate the current change for each device of the same harmonic intensity ratio component from the total current consumption and estimate the movement status for each device. It is possible to grasp the operation status of a plurality of devices even if not. Since it is not necessary to construct a database, it is possible to easily estimate the current consumption, changes, and operating states of individual devices from only the total current consumption values of a plurality of devices at low cost. Moreover, it is not necessary to install a measurement sensor for each electrical device to be measured, just by installing a measurement sensor at an arbitrary point on the customer's power supply line, for example, near the outdoor power line inlet. There is an advantage that privacy is infringed when installed in a house and the degree of additional wiring is low.
[0103]
And grasping | ascertaining the actual usage condition of the electric equipment in such a consumer (a factory, a building, a general household) is important for both an electric business and a consumer. In the case of electric power, it can be utilized for construction of a fee system, development of various energy saving service businesses for consumers, and for consumers for energy saving operation control and detection of equipment failures.
[0104]
Further, according to the inventions of claims 2 and 5, since the current change of the device group having the same harmonic ratio can be further separated based on the information for each device, for example, the current change of the fundamental wave (rated information), The estimation accuracy of the current change amount can be increased.
[0105]
Furthermore, according to the third and sixth aspects of the invention, it is possible to obtain the current consumption and the power consumption for each device group exhibiting the same harmonic ratio.
[Brief description of the drawings]
FIG. 1 is a schematic diagram showing an embodiment of a remote electrical device monitoring method of the present invention.
FIG. 2 is a schematic block diagram showing an embodiment of a remote electrical device monitoring apparatus of the present invention.
FIG. 3 is a flowchart showing an example of an estimation process of a fundamental current change amount for each device group having the same frequency utilization ratio.
FIG. 4 is a flowchart showing an example of an optimal independence index estimation step.
FIG. 5 is a flowchart showing an example of a separation matrix update algorithm (fourth step of the estimation step in FIG. 3) when the optimum independence index G is used.
FIG. 6 is a flowchart showing an example of a process for estimating a current change amount by device from a device group basic current change amount;
FIG. 7 is a graph of changes in effective current and current change of each device (case 197).
FIG. 8 is a graph showing the degree of interference between separation based on effective current of devices and separation based on current changes.
FIG. 9 is a graph showing a current change at each frequency and a current change value of the air conditioner (A1, A2).
FIG. 10 is a graph showing approximate effective current and approximate differential current for each device.
11 is a graph showing the absolute value (average of all cases) of the correlation coefficient between the approximate differential current and the approximate differential current error, where (a) shows the correlation of the approximate differential current and (b) shows the correlation of the approximate differential current error. (The numbers on each axis are device numbers.)
12 is a graph of a signal separation result (case 197) based on a change in effective current by frequency, which is a result of independent component analysis of the present invention. FIG.
FIG. 13 is a comparison diagram of current changes of the first, second, and third independent components and the operating device.
FIG. 14 is a graph showing an operating characteristic model of air conditioner A2 (logarithm of probability of current change).
FIG. 15 is a graph showing a representative independence index and an optimized independence index.
FIG. 16 is a graph showing an estimated current change (in parentheses: corresponding operating device) as a result of independent component analysis using an operation characteristic model;
[Explanation of symbols]
1 Electricity consumers
2 Power supply line
3 Electrical equipment
11 Measuring sensor
12 Frequency component converter
13 time differential device
14 Independent component analyzer
15 Device-specific signal separation device

Claims (6)

In a remote electrical equipment monitoring method for estimating the movement status of a plurality of electrical equipment used by a power consumer, a total load current is measured from a power supply line of the power consumer, and the total load current is measured as a fundamental wave and a harmonic thereof. In addition to conversion to current for each wave, the current difference data is created by taking the time difference between the current for each fundamental wave and harmonics, and the current harmonics for each fundamental wave and harmonics are analyzed by independent component analysis. It is characterized by separating each component estimated as a device group having a ratio, and estimating a movement state (current change) for each device of the monitored device from a waveform of a current change for each component of the same harmonic intensity ratio. Remote electrical equipment monitoring method. Among the current changes for each of the same harmonic intensity ratio components, the component of the device showing the same harmonic intensity ratio is further separated based on the information on the current change intensity of the monitored device to estimate the current change of the individual device. The remote electrical equipment monitoring method according to claim 1. A method for estimating power consumption, wherein the power consumption is estimated from a current change for each device having the same harmonic intensity ratio obtained based on the remote electrical device monitoring method according to claim 1 or 2. . In a remote electrical equipment monitoring apparatus that estimates the movement status of a plurality of electrical equipment used by a power consumer, a total current sensor that measures a total load current from a power supply line of the power consumer, and the total load current A frequency component conversion device for converting the fundamental wave of the total load current and a harmonic current; a time difference device for obtaining a current change by taking a time difference between the fundamental wave of the total load current and the current for each harmonic; and the current An independent component analyzer that separates changes for each component estimated as a device group having the same harmonic intensity ratio by independent component analysis, and a current change waveform for each component of the same harmonic intensity ratio. A remote electrical device monitoring apparatus comprising: a device-specific signal separation device that estimates a movable state (current change) for each device. Among the current changes for each of the same harmonic intensity ratio components, the component of the device showing the same harmonic intensity ratio is further separated based on the information on the current change intensity of the monitored device to estimate the current change of the individual device. The remote electrical equipment monitoring apparatus according to claim 4. 6. A power consumption estimation device characterized in that a power consumption is estimated from a current change for each device having the same harmonic intensity ratio obtained in the remote electrical device monitoring device according to claim 4 or 5.

Images